Under-coding of dementia and other conditions indicates scope for improved patient management: A longitudinal retrospective study of dementia patients in Australia

Abstract

Background:

Under-coding of dementia during hospitalisation results in an inability to identify all patients with dementia using hospital administrative data. Clinical coding can be viewed as a proxy for management; therefore, under-coding indicates dementia was not considered in the patient’s management. While under-coding of dementia is well established, there is sparse evidence on whether dementia is coded in subsequent hospitalisations among patients with a known diagnosis.

Objective:

(a) To describe patterns of dementia coding over 5 years after a first-coded (i.e. index) admission for dementia; (b) to identify factors associated with clinical coding of dementia; and (c) to identify patient subgroups at risk of not being coded to inform future interventions to improve hospital identification and management of dementia.

Method:

Retrospective study of longitudinal hospital data from 1 July 2006 to 30 June 2015 for 7919 patients hospitalised during the 5 years’ post-index admission for dementia in a regional local health district of New South Wales, Australia.

Results:

Dementia was coded in 63.9% of admissions in the 12 months following index admission for dementia; this decreased to 53.7% after 5 years. Patients were 20% more likely to have dementia actively managed when it co-occurred with delirium. Under-coding varied across conditions, with dementia more likely to be coded in admissions for falls and pneumonitis, and less likely for heart failure, pneumonia and urinary tract infection (UTI).

Conclusion:

The frequency with which dementia was not coded highlights opportunities to improve identification and management of dementia through dementia-specific care, enhanced clinical protocols, and interventions focused around heart failure, pneumonia and UTI admissions.

Keywords

dementia clinical coding hospitals health services research health information management

Introduction

Hospital administrative data are widely used in resource allocation, surveillance, epidemiological research, prevention and treatment, and to inform policy (Cheng et al., 2009; Lucyk et al., 2017; World Health Organization, 2018). These data include rich clinical and demographic information, with diagnoses typically classified using the International Classification of Diseases (ICD), which organises and codes health information for the reporting and monitoring of diseases and other health conditions (World Health Organization, 2018). Substantial human resources are dedicated to ensuring the quality, accuracy and consistency of clinical coding of medical information in hospitals (Cummings et al., 2011; Nair, 2013), however, under-coding, particularly of chronic conditions, is prevalent (Peng et al., 2016; Quan et al., 2008). Reasons for this include clinical miscoding resulting from incomplete or unreliable medical documentation, lack of communication between clinical coders and physicians and limitations in the standards and conventions of the clinical coding system (Cheng et al., 2009; Cummings et al., 2011; Lucyk et al., 2017). Recommended strategies to reduce under-coding include increased clinician consultation during the clinical coding process (Mahbubani et al., 2018), and identifying main sources of clinical coding error along the patient trajectory (O’Malley et al., 2005). In Australia, 29 supplementary U-codes for chronic conditions have been trialled since July 2015 to address concerns regarding under-coding of chronic conditions (Australian Consortium for Classification Development (ACCD), 2017). The change directs that chronic conditions (including dementia) should always be coded regardless of whether the condition was managed, even when the condition fails to meet the criteria for clinical coding (as per Australian Coding Standard (ACS) 0002 Additional diagnoses and other clinical coding rules/standards).

Under-coding of dementia during hospitalisation is a significant issue across most Organisation for Economic Co-operation and Development (OECD) countries, which cannot be explained by differences in prevalence (OECD, 2018). A cross-national longitudinal pilot study estimated dementia was not coded in up to a third of admissions among patients with a prior diagnosis of dementia (OECD, 2018), resulting in an inability to reliably identify all people with dementia presenting to hospital. This may influence the quality of care a patient receives, as well as the accuracy of data and the ability to monitor and reduce preventable hospitalisations (OECD, 2018). There is limited evidence using routinely collected data in Australia to assess under-coding of dementia. A single short-term study found that for patients identified with dementia in at least one episode of care in 2006–2007, 47% of their associated episodes were not coded for dementia (Australian Institute of Health and Welfare (AIHW), 2013). These findings are consistent with medical record reviews, which estimate documentation of dementia to be between 53% and 83% (Chodosh et al., 2004; Crowther et al., 2017; Laurila et al., 2004). This is of substantial concern as clinical coding conventions specify that only conditions deemed significant in terms of treatment required, investigations needed and resources used during the admission are coded as diagnoses – thereby providing a measure of patient management (ACCD, 2017; AIHW, 2018). Under-coding of dementia can thus be viewed as proxy for the management of dementia, assuming that if dementia was not coded, it would not have been actively managed during the admission.

The experience of hospital patients with dementia is adversely affected when acute settings fail to provide appropriate services, care and ongoing support (George et al., 2013; Jones et al., 2006; Timmons et al., 2016). The hospital environment can increase agitation, delirium and confusion in patients with dementia (Jones et al., 2006) leading to feelings of disorientation, alienation and loss of control (Digby et al., 2012). When dementia is not identified or actively managed, it may result in adverse inpatient and post-hospital outcomes (e.g. increased length of stay (LOS) and readmissions) and high healthcare costs (AIHW, 2013). Best practice guidelines recommend appropriate and active management of people with cognitive impairment in hospitals, including being alert to delirium and the risk of harm for patients with cognitive impairment, recognising and responding to patients with cognitive impairment and providing safe and high-quality care tailored to the patient’s needs (Australian Commission on Safety and Quality in Health Care (ACSQHC), 2014). This can be brought about using individualised management plans and modifications to the hospital environment to ensure supportive and safe patient care (ACSQHC, 2014).

While under-coding of dementia has been well established (Chodosh et al., 2004; Crowther et al., 2017; Laurila et al., 2004; OECD, 2018), there is sparse evidence on whether dementia is recognised in subsequent hospitalisations among patients with a known admission (principal diagnosis and/or additional diagnosis) for dementia. Previous studies assessing clinical coding following a dementia diagnosis were limited to analysis of a single year (AIHW, 2013), assessed general healthcare utilisation rather than hospital-specific service use (Cho et al., 2014), and did not focus on patterns of clinical coding over time or the impact of clinical coding on patient management or outcomes (AIHW, 2013; Cho et al., 2014; OECD, 2018). Such information is vital to understanding why dementia remains unidentified and unmanaged during admissions, and highlights specific conditions in which to target interventions aimed to improve identification and management of dementia in hospital. This study aimed to describe longitudinal patterns of clinical coding of dementia across subsequent hospitalisations among patients with known dementia by (a) describing rates of dementia coding over the 5 years after a first-coded (i.e. index) admission for dementia; (b) identifying factors associated with lack of clinical coding and, therefore, identification and potential under-management of dementia; and (c) identifying patient subgroups at risk of not being clinically coded to inform future intervention studies to improve identification and management of dementia in hospital.

Method

Design and setting

A population-based retrospective cohort study of hospital utilisation for patients following an index admission for dementia was carried out in a regional local health district (LHD) of New South Wales, Australia. The LHD consists of five hospitals, servicing a regional and rural population of over 400,000 residents (Illawarra Shoalhaven Local Health District, 2018). The study was approved by the Health and Medical Human Research Ethics Committee, University of Wollongong/Illawarra Shoalhaven Local Health District (2017/262).

Data source

Non-identifiable unit-record data were extracted from the Illawarra Health Information Platform’s (IHIP) admitted patient (AP) and emergency department (ED) data collections. This platform consists of a non-identifiable historical databank and a records linkage system for all patients attending Illawarra Shoalhaven LHD facilities (Centre for Health Research Illawarra Shoalhaven Population, 2018). Data were extracted for the study period from 1 July 2006 to 30 June 2015 and a 5-year lookback period from 1 July 2001. Longitudinal data were used as they enable a long-term patient perspective by following individual patients from their first-coded hospital admission for dementia to assess patterns and consistency of dementia coding over time, and to identify when, and under which conditions, under-coding occurred.

Study population

The study population consisted of LHD residents aged 65 years or over at first admission coded with dementia (i.e. index admission) with a hospital separation during the study period. Exclusions were patients aged less than 65 years at index, non-residents and admissions for dialysis (which skewed findings, due to a high volume of separations, and did not present new opportunities for identification of dementia due to lack of routine cognitive screening). Dementia was identified using International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM Ninth Edition) (ACCD, 2017) diagnosis codes of F00 (Dementia in Alzheimer’s disease), F01 (Vascular dementia), F02 (Dementia in other diseases classified elsewhere), F03 (Unspecified dementia), F05.1 (Delirium superimposed on dementia) and G30 (Alzheimer’s disease) (AIHW, 2012) recorded as either the principal or an additional diagnosis. The study consisted of 7920 patients with 33,963 admissions (including index admission and admissions up to 5 years’ post-index). After excluding 2306 episodes for dialysis, the analysis dataset consisted of 7919 persons with 31,657 admissions and 29,978 ED visits.

Matched cohort

A matched cohort was identified by selecting patients with the closest separation date to the index admission for dementia (in the study cohort) with the same age and sex. The matched cohort included ED visits and inpatient admissions at LHD facilities for residents aged 65+ years with no coded dementia-related admission in the 5 years prior to the index admission for dementia, or 5 years’ post-index. The study cohort was matched by exact age for patients 85 years and younger, and grouped for 85+ years (due to small numbers). Once a control was matched, they were excluded from being re-matched. No match was found for two persons. The matched cohort consisted of 8127 patients with 121,002 admissions and 94,268 ED presentations. After excluding 79,171 episodes for dialysis, the matched cohort consisted of 7910 persons with 33,540 admissions and 80,427 ED visits.

Outcome measures

The primary outcome measure was completeness of dementia coding across admissions per year post-index admission. Secondary outcome measures included admission and ED visit patterns by patient characteristics, admission characteristics, diagnosis and principal specialty of care. The diagnostic profiles of the study and matched cohort were compared by major diagnostic category (MDC), principal diagnosis, specialty-related group (SRG) and diagnosis-related group for admissions, and major diagnostic block (MDB) and discharge diagnoses for ED visits. To separate out opportunities for dementia identification, episodes were not concatenated into stays (i.e. in episodes occurring during the same stay) except when measuring patient LOS, which was measured between admission date and final date of separation accounting for transfers.

Statistical methods

Diagnosis was recorded using the ICD-10-AM Ninth Edition (ACCD, 2017) in AP data, and the Systematized Nomenclature of Medicine Clinical Terms Australia and the ICD, Ninth Revision, Clinical Modification in ED data. Potential confounders of the association between dementia and clinical coding analysed included sociodemographic characteristics (sex, age, indigenous status, country of birth, health insurance on admission), co-morbid conditions, patient type (acute, subacute, non-acute, psychiatric), dementia diagnosis type (principal or additional diagnosis) and dementia type. Co-morbidities reported using Ninth Edition ICD-10-AM codes were hypertension (I10–I13, I15), chronic lower respiratory disease (J40–J47), cerebrovascular disease with occlusion/stenosis (I63–I64, I66, G46, I67.8), malignancy (C00–C95), chronic kidney disease (N18), hyperlipidaemia (E78), diabetes mellitus (E10–E14), heart failure (I50), ischaemic heart diseases (I20, I21, I24–I25), arrhythmia (I46, R00) and Parkinson’s disease (G20).

Summary statistics were used to describe outcome variables. ED diagnoses were grouped using the Independent Hospital Pricing Authority’s (IHPA, 2016) Urgency Related Groups (URG) software, URGGrouper (version 1.4.4), to derive MDB. Characteristics of the study population are presented as proportions and percentages for categorical variables, with χ² tests used to assess statistical significance at the 5% level. A Bonferroni correction was applied to adjust significance levels when performing multiple comparisons. Inpatient admissions (N = 803) and ED visits (N = 923) outside of the 5-year window (between 6 years and 10 years of post-index date) were not further analysed, resulting in the difference in size across cohorts. All analyses were performed using SAS version 9.4. (SAS Institute Inc., 2013).

Results

The analysis dataset consisted of 7919 patients with 31,657 admissions and 29,978 ED visits, and the matched cohort included 7910 people with 33,540 admissions and 80,427 ED visits. Missing data rates were low. Dementia diagnosis type was missing in 0.2% of episodes, MDC was missing in 3.2% of the study cohort, SRG was missing in 2.6% of the study cohort.

Patient characteristics

Characteristics at index admission for the study population and matched cohort are presented in Table 1. There were more females than males, and more than 73% of each cohort were aged over 80 years. Patient episodes were primarily acute across both study (89.6%) and matched cohorts (89.8%). On the index admission, dementia was predominantly coded as an additional diagnosis (91.2%). Dementia type was recorded as unspecified in over half (55.4%) of index admissions, with Alzheimer’s disease/Dementia in Alzheimer’s disease the most common specified type. Hypertension was the most prevalent co-morbidity in both cohorts in over 34% of patients, followed by diabetes in approximately one-quarter of patients. The matched cohort had marginally higher rates of co-morbidities, with the exception of Parkinson’s disease, which was higher among those with an index admission for dementia (χ²(1) = 234.45, p < 0.0001). Country of birth was the only patient characteristic which differed across cohorts, with a higher proportion of Australian-born persons in the matched cohort (66.1% vs. 62.9%).

Table 1.

Characteristics at index admission for study population and matched cohort.

Variable	Study population (N = 7919), N (%)	Matched cohort (N = 7910), N (%)	p Value^a
Sex			0.49
Female	4490 (56.7)	4528 (57.2)
Male	3429 (43.3)	3382 (42.8)
Age (years)			0.56
65–69	283 (3.6)	261 (3.3)
70–74	574 (7.3)	532 (6.7)
75–79	1271 (16.1)	1246 (15.8)
80–84	2110 (26.6)	2102 (26.6)
85–89	2292 (28.9)	2329 (29.4)
90+	1389 (17.5)	1440 (18.2)
Aboriginal and/or Torres Strait Islander peoples^b			0.92
No	7857 (99.2)	7853 (99.3)
Yes	50 (0.6)	49 (0.6)
Country of birth^b			<0.0001
Australia	4977 (62.9)	5226 (66.1)
Other English speaking	1149 (14.5)	1133 (14.3)
Non-English speaking	1793 (22.6)	1551 (19.6)
Health insurance on admission^b			0.06
Yes	1809 (22.8)	1679 (21.3)
No	5116 (64.6)	5110 (64.6)
Dementia diagnosis type		NA	NA
Principal diagnosis	695 (8.8)
Additional diagnosis	7224 (91.2)
Dementia type^b		NA	NA
Unspecified dementia	4379 (55.4)
Alzheimer’s disease	1544 (19.6)
Delirium superimposed on dementia	1033 (13.1)
Vascular dementia	586 (7.4)
Dementia in other diseases classified elsewhere	363 (4.6)
Co-morbidities^b^,c
Hypertension	2756 (34.8)	3006 (37.0)	0.0010
Diabetes mellitus	1925 (24.3)	1946 (23.9)	0.7357
Heart failure	1304 (16.5)	1938 (23.9)	<0.0001
Chronic kidney disease	1095 (13.8)	1510 (18.6)	<0.0001
Arrhythmia	844 (10.7)	913 (11.2)	0.0997
Ischaemic heart disease	841 (10.6)	1385 (17.0)	<0.0001
Chronic lower respiratory disease	1249 (15.4)	1249 (15.4)	1.000
Malignancy	716 (9.0)	1567 (19.3)	<0.0001
Parkinson’s disease	607 (7.7)	178 (2.2)	<0.0001
Cerebrovascular disease with occlusion/stenosis	596 (7.5)	554 (6.8)	0.2155
Hyperlipidaemia	230 (2.90)	302 (3.7)	0.0018
Patient type^b			0.06
Acute	7094 (89.6)	7104 (89.8)
Subacute and non-acute	724 (9.1)	786 (9.9)
Psychiatric	101 (1.3)	9 (0.1)

SP: study population; MC: matched cohort.

^aχ² test for SP compared to MC significant at 5% level. Bonferroni correction applied to co-morbidities. χ² test excluded missing, unknown or not stated values.

^b Proportion of missing, unknown, other or not stated: Aboriginal/Torres Strait Islander – SP, 0.2%, MC, 0.1%. Country of birth – SP, 0%, MC, 0.0%. Health insurance on admission – SP, 12.6%, MC, 14.0%. Patient type – SP, 0%, MC, 0.1%. Dementia diagnosis type – SP, 0.2%.

^cCo-morbidities across all admissions (principal and/or additional diagnosis).

Characteristics of hospitalisation (inpatient admissions and ED presentations)

Acute admissions at index and post-index are described in Table 2. Compared to the matched cohort, the study cohort had a higher proportion of admissions for nervous system and musculoskeletal illnesses at index and post-index admission, while the matched cohort was more likely to be admitted with circulatory and respiratory problems. The third most common cause for admission among the study population was diseases and disorders of the kidney and urinary tract. This was higher than the matched cohort both at index and post-index admissions, and was higher at post-index admissions than index admissions for the study population. Subsequent to the index admission, the study cohort were primarily admitted under non-subspecialty medicine (22.9%) and respiratory medicine (12.0%), with 88.6% of admissions being medical, and only 8.8% surgical, less than the matched cohort (16.0%) (χ²(1) = 827.12, p < 0.0001). The study population were more likely to arrive by ambulance and have emergency presentations than the matched cohort, but were triaged to lower categories than the matched cohort (Table 1A in Appendix). The main reasons for ED presentations by the study cohort subsequent to the index presentation were for illnesses of the respiratory (e.g. pneumonia), neurological (e.g. disorientation) and circulatory (e.g. chest pain) systems, and injuries (e.g. fall).

Table 2.

Characteristics of acute patient admissions for SP and MC.

	Index			Post-index
Variable	Study population (N = 7094), N (%)	Matched cohort (N = 7104), N (%)	p Value	Study population (N = 15,948), N (%)	Matched cohort (N = 21,109), N (%)	p Value
Major diagnostic category^a			<0.0001^b			<0.0001^b
Nervous system	1775 (25.0)	545 (7.7)	<0.0001^b	2793 (17.5)	1419 (6.7)	<0.0001^b
Musculoskeletal system and connective tissue	969 (13.7)	806 (11.4)	<0.0001^b	1863 (11.7)	2019 (9.6)	0.0123^b
Circulatory system	782 (11.0)	1350 (19.0)	<0.0001^b	1980 (12.4)	4518 (21.4)	<0.0001^b
Respiratory system	759 (10.7)	856 (12.1)	0.002^b	2000 (12.5)	3082 (14.6)	<0.0001^b
Kidney and urinary tract	543 (7.7)	491 (6.9)	0.01^b	1495 (9.4)	1641 (7.8)	0.0091^b
Digestive system	480 (6.8)	791 (11.1)	<0.0001^b	1264 (7.9)	2283 (10.8)	<0.0001^b
Injuries, poisoning and toxic effect of drugs	316 (4.5)	214 (3.0)	<0.0001^b	868 (5.4)	609 (2.9)	<0.0001^b
Other	1167 (16.5)	1773 (24.7)	<0.0001^b	3181 (19.9)	5039 (23.9)	<0.0001^b
Specialty related group^a,b			<0.0001^b			<0.0001^b
Non-subspecialty medicine	1979 (27.9)	644 (9.1)	<0.0001^b	3645 (22.9)	2396 (11.4)	<0.0001^b
Orthopaedics	755 (10.6)	555 (7.8)	<0.0001^b	1485 (9.3)	1336 (6.3)	0.0050^b
Respiratory medicine	671 (9.5)	757 (10.7)	0.003^b	1907 (12.0)	2842 (13.5)	<0.0001^b
Neurology	666 (9.4)	432 (6.1)	<0.0001^b	1397 (8.8)	1139 (5.4)	<0.0001^b
Cardiology	618 (8.7)	1009 (14.2)	<0.0001^b	1618 (10.2)	3331 (15.8)	<0.0001^b
Gastroenterology	368 (5.2)	540 (7.6)	<0.0001^b	1043 (6.5)	1774 (8.4)	<0.0001^b
Non-subspecialty surgery	347 (4.9)	488 (6.9)	<0.0001^b	1278 (8.0)	1435 (6.8)	0.0026^b
Other	1213 (17.1)	2221 (31.3)	<0.0001^b	3166 (19.8)	6412 (30.4)	<0.0001^b
Diagnosis related group type			<0.0001^b			<0.0001^b
Medical	6227 (87.8)	5258 (74.0)	<0.0001^b	14,131 (88.6)	16,421 (77.8)	<0.0001^b
Surgical	715 (10.1)	1423 (20.0)	<0.0001^b	1397 (8.8)	3386 (16.0)	<0.0001^b
Other	152 (2.1)	423 (6.0)	<0.0001^b	420 (2.6)	1302 (6.2)	<0.0001^b

SP: study population; MC: matched cohort; MDC: major diagnostic category; SRG: specialty related group.

^a Proportion of missing, unknown, other or not stated: MDC – Pre-index: SP, 7.9%, MC, 9.2%. Index: SP, 4.3%, MC, 3.9%. Post-index: SP, 3.2%, MC, 2.4%. SRG – Pre-index: SP, 19.5%, MC, 23.2%. Index: SP, 6.7%, MC, 6.5%. Post-index: SP, 2.6%, MC, 2.1%.

^bχ² test for pre-index, index or post-index admission SP compared to MC significant at 5% level. χ² excluded for missing, unknown or not stated values.

Note: The boldface values are the significance levels for the overall category (ie MDC, SRG, DRG), significant at the 5% level.

Longitudinal reasons for hospitalisation (study population)

Principal diagnoses for acute and subacute admissions for the study population per year post-index are described in Table 3. A peak in both acute (N = 9145) and subacute admissions (N = 6085) occurred in the first year post-index, with urinary tract infection (UTI) the most common principal diagnosis for an acute admission. In the following years, there was an increase in acute admissions for UTI (from 5.2% at index to 6.3% at 4–5 years post-index), and a similar pattern for syncope and collapse (from 1.8% at index to 3.1%). The most common reason for subacute admission was for awaiting admission to residential aged care, particularly in the year post-index (27.3%). Subsequent to the first-coded dementia admission, there was a decrease in acute admissions coded with dementia (from 4.1% to 0.9%) and delirium (from 4.3% to 1.7%). Among subacute patients, there was an increase in dementia clinical coding over the 5-year period for patients awaiting admission to residential aged care (from 17.7% to 25.0%), while a decrease was shown for care involving use of rehabilitation procedure (from 12.0% to 6.1%) and unspecified dementia (from 7.9% to 1.4%). The study cohort had longer acute admissions than the matched for all but two principal diagnoses (pneumonitis and fracture of femur). On subacute admissions, the study cohort also had greater LOS when admitted for all but two principal diagnoses (pneumonitis and pneumonia). Similar to admissions, the first year post-index also demonstrated a peak in ED visits, accounting for almost half (47.8%) of visits to the ED in the 5 years’ post-index (Table 2A in Appendix). Discharge diagnoses for these ED visits were primarily for falls and UTI.

Table 3.

Acute and subacute admitted patient episodes: principal diagnosis for study population by year post-index and year of index.

Years post-index
Principal diagnosis	Index, N (%)	0–1, N (%)	1–2, N (%)	2–3, N (%)	3–4, N (%)	4-5, N (%)	Total, N	ALOS, SP^a	ALOS, MC^b
Acute admissions (N = 23,042 admissions, N = 7094 persons)
Urinary tract infection	368 (5.2)	491 (5.4)	190 (5.7)	99 (5.5)	66 (6.0)	37 (6.3)	1251	13.6	9.8
Pneumonia, unspecified	243 (3.4)	316 (3.5)	156 (4.7)	88 (4.9)	45 (4.1)	26 (4.4)	874	12.1	10.6
Unspecified dementia	294 (4.1)	245 (2.7)	53 (1.6)	24 (1.3)	19 (1.7)	5 (0.9)	640	24.6	–
Pneumonitis (inhalation of food/vomit)	142 (2.0)	194 (2.1)	104 (3.1)	69 (3.8)	36 (3.3)	23 (3.9)	568	11.5	14.1
Delirium superimposed on dementia	303 (4.3)	180 (2.0)	31 (0.9)	25 (1.4)	14 (1.3)	10 (1.7)	563	26.5	–
Congestive heart failure	138 (2.0)	235 (2.6)	88 (2.7)	37 (2.1)	30 (2.7)	10 (1.7)	538	13.7	11.6
Tendency to fall, not elsewhere classified	146 (2.1)	192 (2.1)	68 (2.1)	30 (1.7)	23 (2.1)	10 (1.7)	469	21.9	18.6
Syncope and collapse	130 (1.8)	162 (1.8)	69 (2.1)	31 (1.7)	15 (1.4)	18 (3.1)	425	9.7	6.4
Sepsis, unspecified	141 (2.0)	144 (1.6)	59 (1.8)	34 (1.9)	26 (2.4)	13 (2.2)	417	15.5	13.3
Fracture of subcapital section of femur	157 (2.2)	162 (1.8)	43 (1.3)	29 (1.6)	9 (0.8)	9 (1.5)	409	23.7	27.0
Other	5032 (70.9)	6824 (74.6)	2454 (74.0)	1334 (74.1)	818 (74.3)	426 (72.6)	16888	–	–
Total	7094 (100)	9145 (100)	3315 (100)	1800 (100)	1101 (100)	587 (100)	23042
Subacute admissions (N = 8615 admissions, N = 825 persons)
Awaiting admission: residential aged care	146 (17.7)	1660 (27.3)	182 (22.2)	117 (24.3)	55 (21.7)	37 (25.0)	2197	28.0	19.1
Care involving use of rehabilitation procedure	99 (12.0)	381 (6.3)	45 (5.5)	18 (3.7)	8 (3.2)	9 (6.1)	560	33.1	29.7
Unspecified dementia	65 (7.9)	264 (4.3)	33 (4.0)	18 (3.7)	12 (4.7)	2 (1.4)	394	32.8	–
Awaiting admission to other healthcare facility	21 (2.6)	242 (4.0)	35 (4.3)	16 (3.3)	7 (2.8)	3 (2.0)	324	36.0	32.3
Delirium superimposed on dementia	33 (4.0)	217 (3.6)	18 (2.2)	12 (2.5)	5 (2.0)	6 (4.1)	291	33.7	–
Tendency to fall, not elsewhere classified	21 (2.6)	139 (2.3)	27 (3.3)	16 (3.3)	10 (3.9)	7 (4.7)	220	27.2	21.8
Urinary tract infection, site not specified	11 (1.3)	106 (1.7)	23 (2.8)	7 (1.5)	6 (2.4)	0 (0.0)	153	25.6	24.3
Pneumonia, unspecified	8 (1.0)	95 (1.6)	24 (2.9)	10 (2.1)	4 (1.2)	6 (4.1)	147	17.5	18.6
Congestive heart failure	14 (1.7)	93 (1.5)	15 (1.8)	8 (1.7)	3 (1.2)	2 (1.4)	135	25.4	20.4
Pneumonitis (inhalation of food/vomit)	6 (0.7)	82 (1.4)	24 (2.9)	11 (2.3)	8 (3.2)	3 (2.0)	134	14.1	19.0
Other	401 (48.6)	2806 (46.1)	395 (48.1)	249 (51.7)	136 (53.5)	73 (49.3)	4060	–	–
Total	825 (100)	6085 (100)	821 (100)	482 (100)	254 (100)	148 (100)	8615	–	–
Total admissions (acute and subacute)	7919 (100)	15,230 (100)	4136 (100)	2282 (100)	1355 (100)	735 (100)	31657

^a Average length of stay study population.

^b Average length of stay matched cohort.

Patterns and characteristics of coded admissions

The percentage of admissions coded with dementia for each individual year post-index is shown in Figure 1. Overall, 14,341 (60.4%) of admissions were coded with dementia over the 5 years’ post-index. Clinical coding of dementia was highest in the first year post-index (63.9%, N = 9738), then decreased over the 5 years to 53.7% (N = 395). Presence of delirium was the biggest measured factor impacting on the identification of dementia, increasing dementia coding by 20% in the year post-index (to 83.6%). Following the overall trend, this rate also decreased over time. The same trend was shown for the five most commonly occurring diagnoses (Figure 2), with between 62.8% and 75.7% of admissions coded in the first year post-index, reducing to 52.7–67.6% in the fifth year post-index. Clinical coding rates varied across conditions; the highest rates of clinical coding were when patients were admitted for a fall or pneumonitis, and lowest in admissions with heart failure and pneumonia.

Figure 1.

Percentage of admissions (acute and subacute) coded with dementia by year post-index admission for dementia (N_admissions = 31,657, N_persons = 7919).

Figure 2.

Percentage of admissions (acute and subacute) coded with dementia by condition and year post-index admission for dementia (N_admissions = 31,657, N_persons = 7919).

Discussion

To the best of our knowledge, this study is the first to assess more than 10 years of longitudinal data for patterns of dementia coding in patients after a first-recorded hospital diagnosis of dementia, and only the second Australian study to explore a dataset of more than 2 years, which is significant given the dementia trajectory and how this changes over time (Baker et al., 2017). Our findings provide evidence that dementia remains substantially under-coded over time, with under-coding progressively worsening over a 5-year period following the index hospitalisation. Patients admitted with heart failure, pneumonia and UTI were the least likely to have dementia coded and therefore potentially not actively managed.

Clinical coding of dementia within hospital admissions worsens over time

While the value of consistently identifying and managing a diagnosis of dementia is well established (OECD, 2018), the use of longitudinal data in this study demonstrates that this is not the case over time, with clinical coding of dementia becoming less frequent across subsequent hospitalisations. This is an unexpected result given that dementia is a progressive condition. It may suggest that proximity of subsequent admissions to the index has an impact on clinical coding. Clinical advice is this may be a result of (a) single specialty admissions being less likely to consider dementia as relevant to the admission, (b) the impact of care from the staff’s perspective and (c) therapeutic nihilism (Anderson and Woodburn, 2010; Paterson and Pond, 2009). While further research is required to understand why this is occurring, an important clinical finding is that dementia is not being consistently managed during hospitalisation. As a patient’s dementia progresses, it seems somewhat of a paradox that their likelihood of receiving appropriate care and active management declines – which is of concern for both the management of dementia, and other co-morbid conditions. This highlights opportunities to improve patient care by increasing awareness among specialties outside of geriatrics to consider the patients’ cognitive processes, to methodically collect a collateral history for all elderly patients (Dyer et al., 2018), and to refer patients on when further intervention is required.

Under-coding of dementia in admissions for heart failure, pneumonia and UTI

Dementia was most likely to be under-coded in admissions for heart failure, pneumonia and UTI. These are common conditions for people with dementia (Phelan et al., 2012), however, in this cohort, clinical coding was as low as 39.8% for heart failure, 50.4% for pneumonia and 55.4% for UTI. This consistent under-coding of dementia suggests that patients admitted for these conditions were unlikely to have their dementia actively managed during admission. It is important to understand why this might be occurring to ensure patients with dementia are always receiving appropriate and high-quality care during hospitalisation (ACSQHC, 2014; AIHW, 2013). Clinical advice is that while it is recommended to routinely complete a medical history and check patients’ records for previous diagnoses, this practice is highly variable across different specialties. It is also highly uncommon for clinicians outside of geriatrics to adapt their treatment and management of people with dementia, unless the patient is behaviourally challenging. This is problematic as there is established evidence that modifications to standard care for patients with dementia are beneficial during hospital admissions for heart failure, pneumonia and UTI. Simple nursing interventions can lessen or prevent common complications for UTI, and pneumonia in patients with dementia, including interventions to improve hydration, hygiene and mobility (Bail et al., 2013). Similarly, correct pharmacological management among patients with dementia presenting with heart failure can stabilise or delay cognitive decline (Rozzini et al., 2006; Soto et al., 2013), while specific drugs may impair cognition and cause memory loss (Wagstaff et al., 2003). Active consideration and management of dementia is therefore vital in making appropriate treatment decisions. Failure to do so can lead to adverse outcomes due to compromised ability to manage both the admitting condition and the patient’s dementia, including difficulty complying with dietary restrictions and medications, monitoring fluid and taking suitable action when symptoms of decompensated heart failure develop (Heckman et al., 2007).

The impact of delirium on clinical coding of dementia

Dementia and delirium are intricately linked in the acute setting. Between 30% and 60% of elderly hospital patients have both delirium and dementia (OECD, 2018; Saravay et al., 2004), and prevalence of delirium superimposed on dementia across hospitalised (ED and admissions) patients ranges from 32% to 89% (Fick et al., 2002). This frequent co-occurrence results in issues with differential diagnosis (Jackson et al., 2017). Given the literature, it was anticipated co-occurrence would have an impact on clinical coding rates. This was confirmed and quantified with the finding that patients were 20% more likely to have their dementia coded when it co-occurred with delirium. This may be because dementia becomes more pronounced and requires treatment as a result of the delirium, which significantly accelerates cognitive decline (Fong et al., 2009) and worsens dementia symptoms (Davis et al., 2012). A key clinical implication is that dementia frequently remains unmanaged in hospital except when symptoms are exacerbated as a consequence of co-occurring delirium. Although rates of clinical coding are high in patients with co-occurring delirium in this cohort, they account for one in five admissions. Therefore, while patients with dementia have a higher chance of having their condition managed when it co-occurs with delirium, this accounts for only a small proportion, and may be a consequence of known issues with differential diagnosis.

Distinguishing between dementia and delirium is imperative as assessment and clinical management for each condition is distinct (Fong et al., 2015). The focus of delirium management in hospital is to identify and address the cause, then manage symptoms and prevent complications (National Institute for Health and Care Excellence (NICE), 2010). Survey data suggest this often involves use of pharmacological strategies, particularly when delirium is placing the patient at risk or causing significant distress (Burry et al., 2018; NICE, 2010). Use of antipsychotics, however, can lead to impaired communication and reduced oral intake and ambulation, which can extend the course of delirium (Department of Health and Human Services, 2006) and further exacerbate dementia symptoms. Failure to manage dementia in patients with co-occurring delirium thus has important clinical implications for care and can impact both treatment and assessment. There may also be a relationship between the lack of active management of dementia and the co-occurrence of delirium, whereby earlier and more consistent identification and management of dementia may decrease the likelihood of developing delirium during hospitalisation (Fong et al., 2015; Ford, 2016).

Opportunities to improve management of patients with dementia

The frequency at which dementia is not recognised, and therefore actively managed, highlights the need to differentiate a new care pathway, whereby patients receive dementia-specific care, and clinicians implement enhanced protocols to ensure appropriate management of patients with dementia. This includes an emphasis on the care needs of the patient, including early and appropriate decision-making (Kaplan et al., 2002), and in the case of advanced dementia, consideration of advance directives (Detering and Silveira, 2018).

Implications exist for future planning – particularly when considering the rising numbers of people living with dementia globally, which is projected to increase to 132 million by 2050 (World Health Organization, 2017). In terms of workforce development, hospitals will require a major increase in geriatrician staffing ratios to combat the projected rise in older people with dementia who become hospitalised. Additional training for clinicians outside of the geriatric specialty will also be vital to aid earlier and more consistent identification of dementia, for instance, through simple measures such as ensuring the patient’s medical record is checked for previous diagnoses, and emphasising the availability of structured informant tools to aid comprehensive collateral history taking (Dyer et al., 2018). Further training to enhance clinical ability to distinguish between dementia and delirium is also required to reduce mismanagement of dementia, particularly as diagnosis relies on clinical skills in the absence of a diagnostic test to conclusively determine the cause of cognitive impairment (Young and Inouye, 2007). Hospital interventions to improve clinician ability to recognise and monitor for signs of dementia may benefit from focusing around heart failure, pneumonia and UTI, as these are frequently occurring conditions in which dementia remains unmanaged in hospital in this cohort.

There is cause to improve communication between and within health services to better identify and manage dementia. Previous research demonstrates that patients experience increased frequency of hospitalisation in the year pre-index, primarily for falls and UTI (Cappetta et al., in press). The current study reflects these findings, with falls and UTI among the top reasons for acute admission in the first year post-index, indicating that the patient’s need for additional care was not always appropriately managed, resulting in preventable readmissions for the same condition. Lack of communication between the hospital and the patient’s general practitioner (GP) may be a contributing factor, resulting in the GP missing crucial health information that could better inform care (OECD, 2018). The ability to follow patients longitudinally highlights that greater sharing of diagnostic information and improved quality of communication between hospital and other services at both admission and discharge is required (Hesselink et al., 2012). This sharing could include increased communication and education around services and systems already in place (e.g. understanding the need to refer elderly frail patients with suspected cognitive impairment, and information regarding where to refer them), as well as linking primary care data to other pathways of care (OECD, 2018). Such action is necessary in order to understand and track how people with diagnosed dementia use and navigate health services (OECD, 2018) and to improve the overall care of patients with dementia.

Overall, a fundamental shift is required to improve the identification and active management of patients with dementia. Based on clinical advice and preliminary exploratory analysis, patients admitted under a geriatric specialty demonstrate improved outcomes when compared to other specialties. While this finding requires additional investigation (which the authors are undertaking), it suggests that, ideally, this shift should involve working towards a system where every frail, elderly person with (or without) cognitive impairment is managed in a geriatric service, leaving single-organ specialties to deal with younger populations with single-organ issues.

Limitations of the study

While able to accurately state the frequency and consistency of admissions coded with dementia, we were unable to do the same for clinical coding of delirium. We assumed that if delirium was present, it was identified, as it was impossible to determine otherwise without an audit of medical records, and due to the high prevalence of delirium among elderly hospitalised patients in Australia (Travers et al., 2013). Based on ACS utilised during the study period, it was also assumed that if dementia was coded during the admission, it was managed. This study was also restricted to the use of hospital data, which presents susceptibilities to bias, particularly in terms of under-coding of dementia (Cummings et al., 2011).

To better understand the effect of clinical coding, further investigation using a multivariate model simultaneously accounting for a variety of factors (e.g. patient and admission characteristics) is required to better understand the impact of each factor and interactions between factors, as well as clinical input in interpretation of findings. Further research would be beneficial to understand the implications of under-coding on patient outcomes as more data are collected since the 2015 ICD-10-AM/ACS change which allowed chronic conditions, such as dementia, to be coded when the condition did not meet criteria for clinical coding under ACS 0002 Additional diagnoses (ACCD, 2017). A medical record audit to determine key reasons for under-coding in hospital records would also provide necessary insight. The authors are undertaking such studies as part of a greater body of work. This study also has a number of strengths, including the cohort size and number of years of data, assessment of both inpatient admissions and ED visits and the linkage of patients over time through use of a unique patient identifier.

Conclusion

This study is the first to assess longitudinal patterns of dementia coding in subsequent hospitalisations among patients with a known admission for dementia using more than 10 years of data. Dementia remains consistently under-coded after a first diagnosis in hospital, and progressively worsens across subsequent hospitalisations. This indicates a potential lack of active management of dementia, which was higher for heart failure, pneumonia and UTI admissions. These conditions highlight opportunities to focus and improve interventions at a local level, which target identification and management of patients with dementia in Australian hospitals. It is possible that earlier and consistent identification and active management of dementia may also decrease the co-occurrence of delirium – especially as it relates to hydration, nutrition and use of inappropriate medications in people with dementia.

Footnotes

Acknowledgements

The authors thank Mr Brendan McAllister for support in preparing the data extraction, and Professor Kathy Eagar for assistance with interpretation of findings.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: KC is funded by the NSW Ministry of Health under the NSW Health PhD Scholarship Program. LP is funded under an NHMRC-ARC Dementia Development Fellowship Grant (APP1107401). The authors acknowledge the IHIP research partnership established between the Illawarra Shoalhaven Local Health District (ISLHD) and the University of Wollongong, and ISLHD for funding support and as the source of data used in this study.

ORCID iD

Kara Cappetta

Appendix

Table 2A.

ED discharge diagnosis for study population by year (N = 29,978 visits, N = 5093 persons).

Years post-index
Discharge diagnosis codes^a	Index, N (%)	0–1, N (%)	1–2, N (%)	2–3, N (%)	3–4, N (%)	4–5, N (%)	Total, N
Tendency to fall, not elsewhere classified	447 (8.5)	1213 (8.5)	398 (8.0)	223 (7.8)	147 (8.6)	93 (10.8)	2521
Urinary tract infection, site not specified	235 (4.5)	615 (4.3)	191 (3.9)	115 (4.0)	56 (3.3)	33 (3.8)	1245
Pneumonia, unspecified	132 (2.5)	451 (3.2)	186 (3.8)	94 (3.3)	59 (3.5)	27 (3.1)	949
Fracture of neck of femur, part unspecified	136 (2.6)	418 (2.9)	145 (2.9)	95 (3.3)	40 (2.4)	16 (1.9)	850
Sepsis, unspecified	70 (1.3)	355 (2.5)	146 (2.9)	84 (2.9)	59 (3.5)	23 (2.7)	737
Disorientation, unspecified	114 (2.2)	339 (2.4)	109 (2.2)	36 (1.3)	34 (2.0)	27 (3.1)	659
Unspecified haematuria	26 (0.5)	213 (1.5)	146 (2.9)	171 (6.0)	80 (4.7)	12 (1.4)	648
Other and unspecified abdominal pain	43 (0.8)	273 (1.9)	208 (4.2)	49 (1.7)	16 (0.9)	5 (0.6)	659
Chest pain, unspecified	138 (2.6)	268 (1.9)	105 (2.1)	32 (1.1)	18 (1.1)	21 (2.4)	582
Syncope and collapse	133 (2.5)	227 (1.6)	87 (1.8)	31 (1.1)	23 (1.4)	10 (1.2)	511
Other	3787 (72.0)	9947 (69.5)	3245 (65.3)	1938 (67.6)	1170 (68.7)	595 (69.0)	20,682
Total	5261 (100)	14,319 (100)	4966 (100)	2868 (100)	1702 (100)	862 (100)	29,978

SNOMED CT-AU: Systematized Nomenclature of Medicine Clinical Terms Australia; ICD-9-CM: International Classification of Diseases, Ninth Revision, Clinical Modification; ICD-10-AM: International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification; ED: emergency department.

^a Diagnosis codes mapped from SNOMED CT-AU and ICD-9-CM to ICD-10-AM Sixth Edition.

References

Anderson

Woodburn

(2010) Chapter 2: old-age psychiatry. In: Johnstone

Cunningham Owens

Lawrie

, et al. (eds) Companion to Psychiatric Studies, 8th ed. Edinburgh: Timothy Horne, pp. 635–692.

Australian Commission on Safety and Quality in Health Care (ACSQHC) (2014) A Better Way to Care: Safe and High-Quality Care for Patients With Cognitive Impairment (Dementia and Delirium) in Hospital – Actions for Clinicians. Sydney: ACSQHC.

Australian Consortium for Classification Development (ACCD) (2017) The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM/ACHI/ACS), 10th ed. Darlinghurst: Independent Hospital Pricing Authority.

Australian Institute of Health and Welfare (AIHW) (2012) Dementia in Australia . Report, AIHW, Canberra Australia.

Australian Institute of Health and Welfare (AIHW) (2013) Dementia Care in Hospitals: Costs and Strategies . Report, AIHW, Canberra Australia.

Australian Institute of Health and Welfare (AIHW) (2018) Admitted patient care NMDS 2018–19. Available at: https://meteor.aihw.gov.au/content/index.phtml/itemId/676382 (accessed 19 September 2019).

Bail

Berry

Grealish

, et al. (2013) Potentially preventable complications of urinary tract infections, pressure areas, pneumonia, and delirium in hospitalised dementia patients: retrospective cohort study. BMJ Open 3: e002770.

Baker

Iqbal

Johnston

, et al. (2017) Trajectories of dementia-related cognitive decline in a large mental health records derived patient cohort. PLoS One 12(6): e0178562.

Burry

Mehta

Perreault

, et al. (2018) Antipsychotics for treatment of delirium in hospitalised non-ICU patients. Cochrane Database of Systematic Reviews 6: CD005594. DOI: 10.1002/14651858.CD005594.pub3

10.

Cappetta

Lago

Phillipson

(in press) Patterns of hospital utilisation in the lead up to a diagnosis of dementia: A longitudinal retrospective study of hospital dementia patients in Australia. Australiasian Journal on Ageing.

11.

Centre for Health Research Illawarra Shoalhaven Population (2018) What is CHRISP? Available at: https://ahsri.uow.edu.au/chrisp/what-is-chrisp/index.html (accessed 1 June 2018).

12.

Cheng

Gilchrist

Robinson

, et al. (2009) The risk and consequences of clinical miscoding due to inadequate medical documentation: a case study of the impact on health services funding. Health Information Management Journal 38(1): 35–46.

13.

Cho

Gagnon

Driver

, et al. (2014) Dementia coding, workup, and treatment in the VA New England healthcare system. International Journal of Alzheimer’s Disease 2014: Article ID 821894, 1–5. Available at: https://doi.org/10.1155/2014/821894

14.

Chodosh

Petitti

Elliott

, et al. (2004) Physician recognition of cognitive impairment: evaluating the need for improvement. Journal of the American Geriatrics Society 52: 1051–1059.

15.

Crowther

GJE

Bennett

Holmes

(2017) How well are the diagnosis and symptoms of dementia recorded in older patients admitted to hospital? Age and Ageing 46: 112–118.

16.

Cummings

Maher

Showell

, et al. (2011) Hospital coding of dementia: is it accurate? Health Information Management Journal 40: 5–11.

17.

Davis

Muniz Terrera

Keage

, et al. (2012) Delirium is a strong risk factor for dementia in the oldest-old: a population-based cohort study. Brain 135(9): 2809–2816.

18.

Department of Health and Human Services (2006) Clinical Practice Guidelines for the Management of Delirium in Older People – October 2006. Available at: https://www2.health.vic.gov.au/about/publications/policiesandguidelines/Clinical-Practice-Guidelines-for-the-Management-of-Delirium-in-Older-People---October-2006 (accessed 1 June 2018).

19.

Detering

Silveira

(2018) Advance care planning and advance directives. Available at: https://www.uptodate.com/contents/advance-care-planning-and-advance-directives?topicRef=86250&source=see_link (accessed 20 April 2019).

20.

Digby

Moss

Bloomer

(2012) Transferring from an acute hospital and settling into a subacute facility: the experience of patients with dementia. International Journal of Older People Nursing 7: 57–64.

21.

Dyer

Foley

O’Shea

, et al. (2018) Cognitive assessment of older adults in general practice: the collateral history. Irish Journal of Medical Science 187(3): 683–687.

22.

Fick

Agostini

Inouye

(2002) Delirium superimposed on dementia: a systematic review. Journal of the American Geriatric Society 50(10): 1723–1732.

23.

Fong

Davis

Growdon

, et al. (2015) The interface of delirium and dementia in elderly adults. The Lancet Neurology 14(8): 823–832.

24.

Fong

Jones

Shi

, et al. (2009) Delirium accelerates cognitive decline in Alzheimer disease. Neurology 72(18): 1570–1575.

25.

Ford

(2016) Preventing delirium in dementia: managing risk factors. Maturitas 92: 35–40.

26.

George

Long

Vincent

(2013) How can we keep patients with dementia safe in our acute hospitals? A review of challenges and solutions. Journal of the Royal Society of Medicine 106(9): 355–361.

27.

Heckman

Patterson

Demers

, et al. (2007) Heart failure and cognitive impairment: challenges and opportunities. Clinical Interventions in Aging 2(2): 209–218.

28.

Hesselink

Schoonhoven

Barach

, et al. (2012) Improving patient handovers from hospital to primary care: a systematic review. Annals of Internal Medicine 157(6): 417–428.

29.

Illawarra Shoalhaven Local Health District (2018) Illawarra Shoalhaven Local Health District: about us. Available at: http://www.islhd.health.nsw.gov.au/about_us.asp (accessed 1 June 2017).

30.

Independent Hospital Pricing Authority (IHPA) (2016) URG Grouper Version 1.4.4 User Guide. Darlinghurst: Independent Hospital Pricing Authority.

31.

Jackson

Gladman

JRF

Harwood

, et al. (2017) Challenges and opportunities in understanding dementia and delirium in the acute hospital. PLoS Medicine 14(3): e1002247.

32.

Jones

Borbasi

Nankivell

, et al. (2006) Dementia related aggression in the acute sector: is a Code Black really the answer? Contemporary Nurse: A Journal for the Australian Nursing Profession 21: 103–115.

33.

Kaplan

Angus

Griffin

, et al. (2002) Hospitalized community-acquired pneumonia in the elderly: age- and sex-related patterns of care and outcome in the United States. American Journal of Respiratory and Critical Care Medicine 165: 766–772.

34.

Laurila

Pitkala

Strandberg

, et al. (2004) Detection and documentation of dementia and delirium in acute geriatric wards. General Hospital Psychiatry 26: 31–35.

35.

Lucyk

Tang

Quan

(2017) Barriers to data quality resulting from the process of coding health information to administrative data: a qualitative study. BMC Health Services Research 17: 766–776.

36.

Mahbubani

Georgiades

Goh

, et al. (2018) Clinician-directed improvement in the accuracy of hospital clinical coding. Future Healthcare Journal 5(1): 47–51.

37.

Meagher

Leonard

Donnelly

, et al. (2010) A comparison of neuropsychiatric and cognitive profiles in delirium, dementia, comorbid delirium-dementia and cognitively intact controls. Journal of Neurology, Neurosurgery, and Psychiatry 81: 876–881.

38.

Nair

(2013) Ensuring quality in the coding process: A key differentiator for the accurate interpretation of safety data. Perspectives in Clinical Research 4(3): 181–185.

39.

National Institute for Health and Care Excellence (2010) Delirium: prevention, diagnosis and management. Clinical guideline CG103. Available at: https://www.nice.org.uk/guidance/cg103/resources/delirium-prevention-diagnosis-and-management-pdf-35109327290821 (accessed 1 June 2017).

40.

O’Malley

Cook

Price

, et al. (2005) Measuring diagnoses: ICD code accuracy. Health Services Research 40(5p2): 1620–1639.

41.

Organisation for Economic Co-operation and Development (OECD) (2018) Care Needed: Improving the Lives of People With Dementia, OECD Health Policy Studies. Paris: OECD Publishing.

42.

Paterson

Pond

(2009) The barriers to the early diagnosis of dementia and diagnostic disclosure in primary care. Alzheimer’s & Dementia 5(4): 185.

43.

Peng

Southern

Williamson

, et al. (2016) Under-coding of secondary conditions in coded hospital health data: impact of co-existing conditions, death status and number of codes in a record. Health Informatics Journal 23(4): 260–267.

44.

Phelan

Borson

Grothaus

, et al. (2012) Association of incident dementia with hospitalizations. Journal of the American Medical Association 307(2): 165–172.

45.

Quan

Bing

Saunders

, et al. (2008) Assessing validity of ICD-9-CM and ICD-10 administrative data in recording clinical conditions in unique dually coded database. Health Services Research 43: 1424–1461.

46.

Rozzini

Chilovi

Bertolettie

, et al. (2006) Angiotensin converting enzyme (ACE) inhibitors modulate the rate of progression of amnestic mild cognitive impairment. International Journal of Geriatric Psychiatry 21: 550–555.

47.

Saravay

Kaplowitz

Kurek

, et al. (2004) How do delirium and dementia increase length of stay of elderly general medical inpatients? Psychosomatics 45(3): 235–242.

48.

SAS Institute Inc. (2013) SAS/ACCESS® 9.4 Interface to ADABAS: Reference. Cary: SAS Institute Inc.

49.

Soto

van Kan

Nourhashemi

, et al. (2013) Angiotensin-converting enzyme inhibitors and Alzheimer’s disease progression in older adults: results from the reseau sur la Maladie D’Alzheimer Francais cohort. Journal of the American Geriatrics Society 61: 1482–1488.

50.

Timmons

O’Shea

O’Neill

, et al. (2016) Acute hospital dementia care: results from a national audit. BMC Geriatrics 16: 113–123.

51.

Travers

Byrne

Pachana

, et al. (2013) Delirium in Australian hospitals: a prospective study. Current Gerontology and Geriatrics Research 2013: Article ID 284780, 1–8.

52.

Wagstaff

Mitton

McLendon

, et al. (2003) Statin-associated memory loss: analysis of 60 case reports and review of the literature. Pharmacotherapy 23(7): 871–880.

53.

World Health Organization (2017) Global action plan on the public health response to dementia 2017–2025. Available at: http://apps.who.int/iris/bitstream/handle/10665/259615/9789241513487-eng.pdf;jsessionid=B3C2FF1C2C814379287B6F0265F13179?sequence=1 (accessed 17 June 2019).

54.

World Health Organization (2018) Classifications. Available at: http://www.who.int/classifications/icd/en/ (accessed 1 June 2017).

55.

Young

Inouye

(2007) Delirium in older people. BMJ 334(7598): 842–846.