More than one in two venous thromboembolism treated in French hospitals occurs during the hospital stays

Abstract

Objective

The objective was to describe the prevalence of venous thromboembolism, pulmonary embolism, and deep vein thrombosis among hospitalized patients and the percentages of those occurring during the hospital stays.

Methods

French DRG gave now the opportunity to investigate the frequency of venous thromboembolism occurring during the hospital stay. Statistics are issued from the national PMSI MCO databases encoded using the CIM10. Since 2010–2011 it is possible to differentiate the reason for hospital admission from the pathologies which secondly occurred. Any stay with the ICD-10 codes selected was considered as a hospital-occurred thrombosis unless it was the principal diagnosis of the first medical unit summary. To eliminate outpatient consultations or in day care, stays of <48 h were excluded.

Results

The results pertain to the 78,838,983 hospitalizations in France from 2005 to 2011 and on the 18,683,603 hospital stays in 2010–2011. The incidence of hospital stays came to 860,343 (1.09%) for venous thromboembolism, with 428,261 (0.543%) for deep vein thrombosis without pulmonary embolism and 432,082 (0.548%) for pulmonary embolism. It corresponds to an incidence of 189 per 100,000 inhabitants. Out of 100 hospital stays involving venous thromboembolism, for 40.3% venous thromboembolism was the cause of hospitalization whereas 59.7% can be considered to have occurred during hospital stay. These distributions are of 25.6 and 74.4% for deep vein thrombosis, respectively, 53.8 and 46.2% for pulmonary embolism.

Conclusion

The high proportion of hospital-occurred venous thromboembolism is an alarming situation that should question the quality of prevention and/or its effectiveness.

Keywords

Venous thromboembolism epidemiology hospital-acquired pulmonary embolism

Introduction

Venous thromboembolism (VTE) includes deep vein thrombosis (DVT) and its complication, pulmonary embolism (PE), with a death rate estimated to be in the order of 7–25% depending on the context in which it occurs and the duration of patient follow-up in studies.^1–3 An initial estimate of the incidence of VTE was made as part of a regional prospective study in Brittany (France)⁴ and revealed an overall rate of 183/100,000 inhabitants, subdivided into 124/100,000 for venous thrombosis (VT) and 60/100,000 for PE. More recent estimates have been made using the PMSI MCO (Programme de Médicalisation des Systèmes d’Information—Médecine Chirurgie Obstétrique) national database of summaries of standardized outcomes for evaluating the activity of public- and private-sector hospitals in France and for billing their acts by the health assurance organization. These estimates converge around a similar incidence rate ranging from 187² to 192/100,000 inhabitants in our own research.^5,6 In the United States, this incidence rate was evaluated at 239/100,000 for the same period. These incidence rates in the population probably underestimate the true prevalence of these pathologies because many instances of DVT go unnoticed as the symptoms are not very apparent while many deaths from PE go undiagnosed because autopsies are not systematically performed, especially on the elderly who die at home or in nursing facilities.

In this study, which also uses the national PMSI MCO database, we focus on the incidence of VTE not in the French population but with respect to the number of hospital admissions in France. This provides a precise picture of distribution by age and sex of the patients admitted but the major advantage of the approach is that, for the first time in France, a distinction can be made between VTE as a cause of hospitalization and VTE which are not the reasons for which patients are admitted to hospital and occurred during the hospital stay.

These incidence rates are also reliable and reproducible indicators for monitoring any progress in the prevention of thromboembolic diseases both in primary care and hospital practice.

Methodology

Hospitalization data

The statistics are from the national PMSI MCO databases inspired by the US Medicare system. They are compiled each calendar year from the RSA (anonymous discharge summary) files forwarded and validated by health establishments with admissions in medicine, surgery, obstetrics, and odontology (MCO). The anonymous summaries are encoded using the 10th edition of the international classification of diseases. The codes used for characterizing VTE are I801–I809 for DVT and codes I260, I269 for PE. The analyses identify all VTE, DVT without PE (DVT), and PE with or without previous/associated diagnosis of DVT. The study data cover the period 2005–2011. For 2010 and 2011, stays have as the Principal Diagnosis of Medical Unit Summaries the reason for admission to the medical unit and no longer the diagnosis that mobilized most resources of the medical unit. This development makes it possible to distinguish reasons for which patients are admitted and the pathology which secondly occurred during the stay. Any stay with a mention of one of the ICD-10 codes selected regardless of the Principal Diagnosis of Medical Unit Summaries and whatever its position (Principal Diagnosis, Related Diagnosis, Significant Associated Diagnosis) was taken to have occurred unless it was the Principal Diagnosis of the first Medical Unit Summary of the stay. In fact, if VTE diagnosed at inclusion was considered to be the cause of hospitalization, this condition was coded as the Principal Diagnosis of the first Medical Unit Summary of the stay. In order to center the study on hospital stays and eliminate any diagnoses made at outpatient consultations or in day care, any stays of less than 48 h were excluded from the analyses. Therefore, this study does not provide an evaluation of the VTE prevalence and especially of DVT in France because many patients with DVT are treated in community practice without requiring to be hospitalized but gives the prevalence of VTE among hospital stays and the proportion of them that appeared during the hospital stays and were not the reason for which the patient was initially hospitalized.

Statistical analyses

The descriptive results are presented as percentages for qualitative variables and by means and standard deviations for quantitative variables. Means are compared using Student’s t-tests and percentages using Chi² tests. The software used is SAS version 9.3.

Results

The results pertain to the 78,838,983 hospitalizations in France from 2005 to 2011, namely 11,504,517 in 2005, 10,933,975 in 2006, 10,867,379 in 2007, 10,927,019 in 2008, 11,294,130 in 2009, 11,621,958 in 2010, and 11,690,005 in 2011.

Study of the overall hospital incidence

Data from the national database reveal that over the period 2005–2011 the incidence of hospital stays came to 860,343 (1.09%) for VTE, with 428,261 (0.543%) for DVT without PE and 432,082 (0.548%) for PE. All told, the mean number of VTE hospitalized per year over the period was 122,906, including 61,180 for DVT and 61,726 for PE. Out of the French national population in 2010, those 122,906 admissions for VTE correspond to an incidence of 189 per 100,000 inhabitants. These instances of VTE occurred for 43.4% in men and 56.6% in women. The percentages are, respectively, 42.71 and 57.29% for DVT and 44.09 and 55.91% for PE. The mean age of VTE patients is 69.1 ± 16.8 years, 68.9 ± 17.0 for DVT and 69.3 ± 16.8 for PE. Table 1 shows the annual incidences for 100 hospital stays for 2005–2011 and the aggregate total over all those years and reveals a remarkably stable incidence over the seven years monitored.

Table 1.

Annual incidence for 100 hospital stays.

	2005	2006	2007	2008	2009	2010	2011	2005–2011
VTE	1.115	1.107	1.102	1.099	1.088	1.076	1.054	1.091
DVT	0.575	0.571	0.556	0.545	0.534	0.521	0.504	0.543
PE	0.540	0.537	0.547	0.553	0.555	0.555	0.550	0.548

The incidence of VTE, DVT, and PE by sex and age at the time of hospitalization is shown in Tables 2 and 3. Table 2 reveals a slightly higher incidence (+0.1%) of diagnosis of VTE among causes of hospitalization for women than for men. Table 3, by contrast, shows that, for patients aged 60 and more, hospitalizations for VTE and DVT are nearly four times (3.87 and 3.73) higher than diagnosis in the population aged less than 60 years and more than six times (6.1) higher for PE. These incidence rates remain very steady throughout the years covered by the study.

Table 2.

Annual incidence for 100 hospital stays by sex.

	Sex	2005	2006	2007	2008	2009	2010	2011	2005–2011
VTE	M	1.033	1.024	1.040	1.038	1.037	1.018	1.012	1.029
VTE	F	1.184	1.177	1.155	1.150	1.133	1.128	1.091	1.145
DVT	M	0.526	0.522	0.518	0.506	0.501	0.483	0.475	0.504
DVT	F	0.616	0.612	0.588	0.579	0.562	0.555	0.529	0.577
PE	M	0.508	0.502	0.522	0.533	0.536	0.535	0.536	0.525
PE	F	0.567	0.566	0.567	0.571	0.571	0.573	0.562	0.568

Table 3.

Annual incidence for 100 hospital stays by age.

	Age	2005	2006	2007	2008	2009	2010	2011	2005–2011
VTE	<60	0.483	0.480	0.482	0.490	0.496	0.498	0.496	0.489
VTE	≥ 60	2.048	2.006	1.961	1.922	1.845	1.791	1.728	1.894
DVT	<60	0.264	0.264	0.252	0.251	0.245	0.241	0.235	0.250
DVT	≥ 60	1.034	1.011	0.976	0.944	0.903	0.868	0.829	0.934
PE	<60	0.145	0.140	0.150	0.158	0.166	0.171	0.019	0.158
PE	≥ 60	1.014	0.996	0.985	0.978	0.943	0.923	0.899	0.960

Distribution of VTE hospital occurred or not

The results bear on the 23,311,963 hospital stays of more than two days in 2010 and 2011 during which the information collected can now be used to identify patients admitted for VTE and those VTE occurred during the stay, i.e. admitted for another reason.

Analysis shows that out of 100 hospital stays involving VTE, for 40.3% VTE was the cause of hospitalization whereas 59.7% can be considered to have occurred during hospital stay. These distributions are of 25.6 and 74.4% for DVT and the ratio is reversed for PE with 53.8% being the cause of hospitalization and 46.2% being admitted for another reason. Table 4 shows that the age of patients varies little with whether VTE, DVT, and PE were the cause of hospitalization, or not and are similar in men and women.

Table 4.

Mean age (± SD) and distribution by sex of VTE hospital occurred or not.

%	VTE		DVT		PE
%	Cause of hospitalization	Hospital occurred	Cause of hospitalization	Hospital occurred	Cause of hospitalization	Hospital occurred
Age	67.22 ± 18.11	69.65 ± 16.34	67.90 ± 17.83	70.46 ± 16.23	68.13 ± 17.41	69.91 ± 16.19
Men	40.1	59.9	26.1	73.9	52.4	47.6
Women	40.5	59.5	25.2	74.8	54.9	45.1

Table 5 shows the frequency of VTE, DVT, and PE whether the cause of hospitalization or not depending on patient age. It reveals that regardless of age, the proportion of patients presenting DVT as the cause of hospitalization is lower than the proportion for whom DVT was not the cause of hospitalization, whereas the reverse is observed for PE. All told, VTE is less frequently the cause of hospitalization because of the higher frequency of DVT than PE. It can be observed that although PE as the cause of hospitalization is more frequent than PE which was not the cause of hospitalization, across all age groups the percentage of hospital-occurred PE is between 36.3 and 47.5%. It can also be observed that the percentage of hospital-occurred VTE, DVT, and PE is very high in the under-18 years age group even if it is happily relatively rare, that it then falls between the ages of 20 and 50 before rising and leveling off beyond that age for all age groups, including the 80 years and more group.

Table 5.

Distribution by percentages of VTE, DVT, and PE hospital occurred or not by patient age.

%	VTE		DVT		PE
%	Cause of hospitalization	Hospital occurred	Cause of hospitalization	Hospital occurred	Cause of hospitalization	Hospital occurred
<18 years	39.2	60.8	31.3	68.7	52.5	47.5
18–39 years	53.2	46.8	40.2	59.8	63.7	36.3
40–49 years	47.8	52.2	34.4	65.6	59.5	40.5
50–59 years	40.0	60.0	26.3	73.7	52.4	47.6
60–69 years	38.3	61.7	22.0	78.0	52.9	47.1
70–79 years	38.4	61.6	22.3	77.7	52.5	47.5
80 years and more	38.9	61.1	25.0	75.0	52.5	47.5

The percentage of mortalities during a hospital stay from all these cases of VTE is high and reaches 6.58% and the mortality from VTE, DVT, and PE is multiplied by a factor of 3–4 (p < .0001) when hospital occurred (Table 6).

Table 6.

Percentage of deaths during hospital stays depending on whether the various forms of VTE were hospital occurred or not.

	VTE			DVT			PE
	Cause of hospitalization	Hospital occurred	Total	Hospital occurred	Cause of hospitalization	Total	Cause of hospitalization	Hospital occurred	Total
Deaths %	2.99	8.96	6.58	1.98	7.46	5.34	3.95	14.01	8.60

Discussion

This study of the incidence of VTE in hospitals and of whether or not it is hospital occurred has certain methodological limitations.

The first limitation lies in the quality of VTE coding by the ICD-10 in the PMSI-MCO context. One study has shown its sensitivity was high (89%) for PE encoding but comparatively low (58%) for DVT, which might contribute to underestimating its incidence.⁷ Conversely, it may be that the difference between superficial and deep phlebitis is not consistently observed and that on the contrary certain instances of superficial phlebitis of the upper limbs might be encoded I808 or I809, that is “Phlebitis and thrombophlebitis of other sites or unspecified sites” thus augmenting the DVT. Vice versa, PE may be underestimated in this study. Work on 1000 patients including an autopsy study found the cause of death in 15.9% of cases to be PE.⁸ Another autopsy study⁹ reveals that only 45% of deaths by PE had been diagnosed before death. It may be therefore that our study overestimates DVT a little and underestimates PE.

However, the incidence of VTE of 189 per 100,000 inhabitants reported in this study is comparable with the 183 for 100,000 inhabitants described in the study by E. Oger⁴ for the Brittany region in 2000 and the 187 per 100,000 inhabitants reported by V. Olie² for the national population PMSI MCO for 2010 and which a priori contains the same methodological bias. It is worth pointing besides to the great stability of figures in the annual incidence 10 years apart, which is also apparent for all seven years of our annual monitoring from 2005 to 2011.

The study can also be criticized for failing to distinguish between the initial episodes and relapses or recurrences of VTE. This bias may affect the epidemiological incidence of VTE in the general population but not the incidence of VTE among diseases for which care is delivered in hospital nor the fact that instances of VTE were the initial cause of hospitalization or hospital occurred.

One other limitation is the term of hospital occurred that we use for simplifying the text and the real term should be not officially the reason for hospitalization and in some cases some VTE could have been already present at hospital admission in parallel, for example to a very important renal failure due to a kidney cancer. In such a case, the primary diagnosis which has been recorded is surely kidney cancer. It is also important to note that no systematic risk assessment of VTE is used when the patient is hospitalized and that Nice quality standards are far from being used everywhere. Today, practitioners are more focused on an early diagnosis of VTE but not on its systematic prevention and that perhaps explain the high rate occurring during the hospital stay.

Our findings reveal a slightly higher incidence (+0.1%) of VTE diagnosed as the cause of hospitalization among women than men. This is a small difference and reflects the fact that such incidences vary among women with age. We did not conduct the analysis as a function of age and sex simultaneously but it is likely that, as in the study by Olie, VTE is far higher among women of childbearing age² than in men because of hormonal factors during this period of women’s lives.¹⁰ This trend probably reverses subsequently because of many other causes of hospitalization among women whereas, among men, diagnoses of cardiovascular disease and cancer become predominant. Conversely, very large variations are observed as a function of age. The frequency of diagnosis for hospitalization for VTE and DVT is almost four times (3.87 and 3.73) higher among patients aged over 60 years than in the population aged less than 60 years and more than six times higher (6.1) for PE. These increases in terms of diagnoses for hospitalization are consistent with the MMWR data¹¹ in the United States which show for VTE rates of 82 and 111 for 100,000 for the 40–49- and 50–59-year age groups, rising to 203 for 60–69 years, 349 for 70–79 years, and 500 after 79 years. They are also consistent with epidemiological data showing an increased incidence of VTE in the population as a function of age.¹²

Whether for variations with sex or age, there is also great stability of the rates of incidence throughout the years covered by the study.

But beyond this confirmation of the high incidence of VTE whether in terms of incidence within the population or among hospitalization diagnoses, the main finding of the study is to highlight that 60% of cases of VTE cared for in hospital was not the cause of hospitalization but was hospital occurred. As for PE, the prognosis for which is particularly severe, the percentage is admittedly lower but 40% of occurrence during a hospital stay is a particularly worrying figure.

In methodological terms, the choice to transpose the 48-h limit used for nosocomial infections to hospital-occurred VTE is debatable but it does provide a relatively well-established reference^6,13 and has already been transposed to other spheres, particularly bed sores¹⁴ or other undesirable complications arising from hospitalization which share the feature of being potentially avoidable. Potentially avoidable, of course, because the prevention of VTE like the prevention of other infections or bed sores can only reduce their incidence and not eradicate them completely.

We do not say that these VTE occurring during the hospital stay are “nosocomial” but however, the fact that six out of 10 instances of VTE or four out of 10 cases of PE occur in the course of a hospital stay is particularly high, and in any event too high to be compatible with properly conducted prevention with current means. One might look especially at the use of elastic compression stockings which are generally available in hospitals that fail to exert the required pressure to ensure effective prevention (minimum 15–20 mmHg) and the still limited use of anticoagulants whether of the low-molecular-weight heparine type, fondaparinux, or new oral anticoagulants the use of which is reserved primarily to certain particularly thrombogenic postsurgical situations.

One hypothesis that still remains to be tested in the available data base is that prevention may be actively put in place for some pathologies where it has truly achieved the status of medico-legal obligation as in orthopedic or gynecological surgery but is still too often neglected for hospitalization of the elderly for downturns in their general condition or infectious or rheumatological syndromes and who are exposed to a major risk factor, namely bed rest. However, the French health authorities have issued recommendations.¹⁵ Another area where venous thromboprophylaxis is not developed enough is probably cancer, although it is known that VTE is frequently a complication of this illness and recommendations have been made internationally.^16,17

Another worrying feature highlighted by our study is the higher death rate from hospital-occurred VTE and PE than when VTE and PE are the initial causes of hospitalization. True, the situation may be explained by the fact that the conditions occurred in subjects who were probably in a poor overall state of health and that the data available did not allow the diagnosis of the cause of death to be identified in the PMSI MCO. It can also be suggested that, unlike in people for whom the diagnosis for hospitalization was VTE, VTE occurring in subjects hospitalized for other reasons was less often identified and actively cared for. This hypothesis is unfortunately consistent with autopsy results showing that in 45% of patients who died of PE, PE had not been diagnosed before the autopsy.⁹ This points out the potential benefit for the patients which could result from a systematic application of the NICE Quality Standard for VTE prevention and some work recently presented by our team showed that it will be cost-efficient.¹⁸ It is today unknown if the higher death rate of PE occurring at hospital is linked to the associated comorbidities. That assumption may be reasonably proposed as an explanation but VTE has also a strong lethal effect.

Subsequent work will attempt to explore these avenues with as their limits that the PMSI discharge summaries do not constitute a medical dossier summary but a collection of information for the purposes of evaluating hospital activity and its billing to the payer organizations. However, the indicators arising from it, such as those in this study, have the advantage of being exhaustive and reproducible. This makes them valuable for quantifying advances that might be made by bolstering thrombosis prevention policies. Such a VTE prevention policy would have advantages not just in terms of patient benefits but also in terms of health savings because cover for VT or PE is expensive for the health assurance organization both in terms of extra hospital costs implied and of the future monitoring of the patient by the primary care practitioner. Economic approaches are under study based on hospital data for evaluating the cost of such hospital-occurred VTE and putting in perspective the costs of prevention of VTE and of care delivery for VTE. As for patient benefits, the stakes are high: INSERM data on all information from death certificates put the number of deaths in France related to the occurrence of VT at 20,000 per year. In the light of our findings, a large proportion of these deaths are probably due to hospital-occurred VTE and avoidable. This poses for hospitals the question the present authors raised nearly 20 years ago about prevention in primary care medicine: Are we doing enough to prevent DVT?¹⁹

This situation is probably not confined to French public- or private-sector hospitals but we have not found any publications about the hospital-occurred character or otherwise of VTE for other countries. The marked parallel of frequencies reported in hospitals in France and the United States in the MMWR publication⁷ suggests the same might apply.

Conclusion

VTE alone makes up nearly 1% of all hospital stays and the high proportion of hospital-occurred VTE is an alarming situation that should make us question the quality of prevention implemented and/or its effectiveness. VTE prevention policies must be strengthened in hospitals for the sake of patients and healthcare savings alike. It would be easy to monitor results from the simple indicators developed in this paper that provide a benchmark for evaluating future improvements.

Footnotes

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by grants from University Hospital of Dijon.

References

Goldhaber

Visani

De Rosa

. Acute pulmonary embolism: clinical outcomes in the International Cooperative Pulmonary Embolism Registry (ICOPER). Lancet 1999; 353: 1386–1389.

Olié

Chin

Lamarche-Vadel

. La maladie veineuse thromboembolique: patients hospitalisés et mortalité en France en 2010. Bull Epidémiol Hebd 2013; 33–34: 417–424.

Meyer

Sanchez

. Pulmonary embolism. EMC-Anesthésie Réanimation 2004; 1: 54–68.

EPI-GETBP Study Group. Groupe d’Etude de la Thrombose de Bretagne Occidentale Incidence of venous thromboembolism: a community-based study in Western France. Thromb Haemost 2000; 83: 657–660.

Allaert

Benzenine

Quantin

. Prévalence de la maladie thromboembolique veineuse (MTEV) dans les hôpitaux de France et des Etats-Unis. Angéiologie 2014; 65/66: 30–31.

Allaert

Elalamy

Benzenine

. Prévalence hospitalière des thromboses veineuses profondes et des embolies pulmonaires constituant le motif d’admission ou acquises à l’hôpital. Angéiologie 2012; 63: 50–51.

Casez

Labarère

Sevestre

. ICD-10 hospital discharge diagnosis codes were sensitive for identifying pulmonary embolism but not deep vein thrombosis. J Clin Epidemiol 2010; 63: 790–797.

Pineda

Hathwar

Grant

. Clinical suspicion of fatal pulmonary embolism. Chest 2001; 120: 791–795.

Kakkar

. Pulmonary embolism in medical patients: an autopsy-based study. Clin Appl Thromb Hemost 2008; 14: 159–167.

10.

Middeldorp

. Thrombosis in women: what are the knowledge gaps in 2013? J Thromb Haemost 2013; 11: 180–191.

11.

Centers for Disease Control and Prevention. Venous thromboembolism in adult hospitalizations; United States, 2007–2009. Morb Mortal Wkly Rep 2012; 22: 402–403.

12.

White

. The epidemiology of venous thromboembolism. Circulation 2003; 107: I4–8.

13.

CTINILS . Définition des infections associées aux soins. Ministère de la santé, de la jeunesse et des sports 2007, pp. 40–41.

14.

Allaert

. L’escarre comme affection nosocomiale. L’escarre 2001; 11: 25–25.

15.

AFSSAPS. Prévention et traitement de la maladie thromboembolique veineuse en médecine, 2009: 3–5.

16.

Prandoni P, Piovella C, Filippi L, et al. What are the pharmacotherapy options for treating venous thromboembolism in cancer patients? Expert Opin Pharmacother 2014; 15: 799–807.

17.

Farge

Debourdeau

Beckers

. International clinical practice guidelines for the treatment and prophylaxis of venous thromboembolism in patients with cancer. Thromb Haemost 2013; 11: 56–70.

18.

Allaert

Benzenine

Quantin

. Fréquence et coûts des thromboses veineuses profondes et des embolies pulmonaires survenant au cours du séjour hospitalier. 73ème Journées de la Société Française de Phlébologie. Phlebologie 2014; 4: 69–70.

19.

Allaert FA, Vin F and Shadeck M. Are we preventing enough deep venous thrombosis in medical patients. In: European Congress of the International Union of Phlebology, Budapest, 6–10 September 1993. Abstract Book: 78.