Tuberculosis and the Acute Abdomen: An Evaluation of the National Inpatient Sample

Methods

We used data from the Agency for Healthcare and Quality, Healthcare Cost and Utilization Project's (HCUP) National Inpatient Sample (NIS) from 2010–2014 to describe the epidemiology of acute abdominal pathology associated with Tb [11]. The NIS is the largest all-payer inpatient database in the United States, covering more than 96% of the US population [12]. Publically available, the NIS is constructed from the State Inpatient Databases and samples approximately 20% of US hospitals [12]. When analyzed, the NIS is weighted to provide national estimates of inpatient admission. The number of hospitals included in the sample ranged from 1051–4411, and reported 7,071,762–8,023,590 (35,358,818–37,352,013 after weighting) annual inpatient encounters over the five-year period [12].

We included all inpatient encounters with the diagnosis code corresponding to infection with Tb (International Classification of Disease, 9th Revision, Clinical Modification [ICD-9CM] 010-018). We excluded ICD-9 CM code 0164 because of misclassification as Mycoplasma genitalium infection. A diagnosis of acute abdominal pain was defined as an admission associated with ICD-9 CM codes: 789.0, 789.4, 789.59, 789.6, 789.7, 789.9, 567.21, 567.22, 567.23, 567.29, 567.31, 567.38, 567.39, 567.81, 567.82, 567.89, 567.9.

Acute abdominal Tb was defined as an ICD-9 code for Tb, and an ICD-9 code for acute abdominal pain. This was done in an effort to capture cases that may be initially coded as other forms of Tb but have acute abdominal pathology, representing un- or underreported abdominal complications associated with Tb.

Patient variables assessed included age, gender, race, US census region, payer status, income quartile, comorbid conditions, alcohol and drug use, HIV or acquired immune deficiency syndrome (AIDS) status, length of stay, death, and discharge destination. Comorbid conditions were assigned based on the Elixhauser Comorbidity Algorithm (Elixhauser Comorbidity Software, Version 3.7) according to existing data elements and coding conventions [13,14]. The 2010–2014 period was chosen because these databases were both available at the time of analysis and had consistent coding to enable combinatorial analysis. Outcomes of interest included need for operation and death.

All statistical analyses used the NIS sampling strata and discharge weights to produce nationally weighted patient-level estimates and 95% confidence intervals (CIs) that account for clustering of patients among hospitals. Weighted values are reported unless otherwise specified. Uni-variable and multi-variable analyses were performed. The a priori p value considered significant on uni-variable analysis for inclusion into multi-variable analysis was p < 0.1; a p < 0.5 was considered significant on multi-variable analysis.

Fisher exact tests were used where appropriate. Cell contents with less than 10 cases are reported as <10 in accordance with HCUP user agreement [12]. Stata 12.0 (StataCorp, College Station, TX) was used for all statistical analyses. The study was classified exempt after Institutional Review Board review because it contained no identifiable data.

Results

There were 66,034 inpatient admissions associated with Tb infection of which 3638 (6%) had acute abdominal pain meeting our case definition (Table 1). Among these, 1578 (43%) were 45–64 years old and 2344 (64%) were male. Most patients were Hispanic (n = 1090, 30%) or black (n = 924, 25%). Four hundred and eighty-nine (13%) patients had a concomitant diagnosis of alcohol abuse or drug abuse (n = 271 [7%]). One hundred and twenty-five patients had a codiagnosis of HIV or AIDS. Weight loss (n = 1057, 29%) was common among patients.

More cases were reported from the South (n = 1321, 36%) and West (n = 1169, 32%) regions. Medicaid (n = 1195, 33%) was the most common payer source. The lowest quartile of income by zip code (n = 1367, 38%) had the most patients. Most patients had primary Tb diagnoses of pulmonary (n = 1928, 53%), miliary (n = 522, 14%), or abdominal (n = 482, 13%) Tb.

A total of 347 (0.5% of total, 10% of cases with acute abdominal pain) patients underwent a surgical procedure. Of those, 111 (32%) underwent an abdominal exploration. Other procedures included peritoneal biopsy (n = 136, 39%), repair or resection of a hollow viscus (n = 122, 35%), lysis of adhesions (n = 88, 25%), and a procedure involving the liver, gallbladder, or pancreas (n = 44, 13%). The single most common operation was a peritoneal biopsy (n = 134, 39%).

In adjusted analysis, an association between a diagnosis of Tb and undergoing a surgical procedure was found to depend on the type of Tb infection (odds ratio [OR] = 1.17 for intestinal, peritoneal, or genitourinary Tb vs. other types, 95% CI = [1.12–1.22]) and whether the patient was of white, Asian, or other race versus black and Hispanic (OR = 1.11, 95% CI [1.02–1.21]) (Table 2). The significance of race on undergoing operation remained significant on hierarchical regression models with minimal additional variance from socioeconomic indicators (Table 3).

Thirty-nine (11%) of the 347 patients who underwent a surgical procedure died during hospitalization (Table 4). Patients who died during hospitalization were older on uni-variable analysis, with 49% of patients 65 years or older (p = 0.04). Patients who were black or Hispanic or other races were more likely to die than their white or Asian counterparts, although the difference did not meet statistical significance. The frequency of all comorbidities was greater among patients who died, although only hypertension (64% vs. 22%, p = 0.03) and coagulopathy (49% vs. 11%, p = 0.02) reached statistical significance on uni-variable analysis. The payer source for all patients who died was either Medicare or Medicaid (p = 0.04).

All patients were hospitalized for at least one week before they died. After controlling for known confounders, increased death after operation was associated with concomitant hypertension (OR = 12.11, 95% CI [1.33–110.00]) and coagulopathy (OR = 22.56, 95% CI [2.46–207.08]) (Table 5).

Discussion

In the United States, infection with Tb is more commonly associated with the foreign-born, necessitous, and HIV-infected populations [6]. Foreign-born individuals emigrating from countries with high endemic Tb rates accounted for 70% of US cases in 2017 [4]. Because of substantial public health efforts, the US Tb incidence has decreased over the past several decades, falling to a historic low in 2017 at 2.8 per 100,000 [4].

While both pulmonary and EPTb rates have continued to decrease, the proportion of EPTb compared with total Tb rates has increased over time, from 8% of cases in 1962 [15] to 16% of cases in 1993 [16] to 21% of cases in 2017 [4]. This phenomenon may be explained partially by the observed influx of foreign-born persons (e.g., Southeast Asians) who are known to have higher prevalence of EPTb than the general Tb population [7]. Yet, this trend has also been seen in other developed countries that do not have the same proportion of foreign-born cases [8]. Other possible etiologies include the emphasis placed on prevention of pulmonary Tb by existing public health protocols, and that EPTb may be poorly recognized because of its non-specific symptoms [7].

The 6% abdominal Tb rate we identified in NIS is consistent with previous assessments of Tb surveillance, which have estimated abdominal Tb rates of 3%–12% [4,7,8]. Similar to the relationship between EPTb and all Tb cases, the proportion of peritoneal Tb compared with EPTb has also been rising slowly from 5% in 2011 to 6% in 2017, although this increase between studies is likely insignificant [4]. Understanding the true rate of change of abdominal Tb is difficult. Accurate diagnosis remains a challenge; there may be underdiagnosis because of non-specific symptoms or overdiagnosis because a minority of cases are confirmed by culture [8].

The epidemiologic attributes of Tb patients with acute abdominal pain were similar to those of the general Tb population—the majority were older (45–64 y: 43%), male (64%), and of lower socioeconomic class (38% in lowest quartile of income by zip code) [1,4]. When analyzing patients with a diagnosis of acute abdominal pain who underwent a surgical procedure, however, we identified differences between expected demographics and those observed in NIS. Race in particular was discrepant.

In previous analysis of surveillance datasets, most US EPTb cases occurred in nonwhite racial and ethnic groups, and blacks typically comprised the largest group [7]. While these racial trends were also present in the NIS dataset, a disproportionate number of whites and Asian patients underwent abdominal operation. Possible explanations include either white or Asian patients tend to have more severe disease, an increased propensity for complications to develop, have unequal access to surgical care, or face surgeon bias when undergoing evaluation. We suspect the latter two explanations are more likely. White or Asian patients may be more likely offered operation or are in a healthcare setting where a surgical procedure is an option. This hypothesis is supported by previous studies suggesting provider bias may lead to differential treatment based on ethnicity or race [17,18].

Definitive diagnosis of abdominal Tb is made by demonstration of M. tuberculosis in peritoneal fluid, biopsy specimen, or culture/nucleic acid amplification test [7,19,20]. The peritoneum, intestine, and liver are common sites of disease, although any abdominal organ or space may be involved [2,21,22]. In the NIS dataset, the most common surgical procedure was a biopsy, suggesting the persistent utility of operation as a diagnostic tool. Operation may also be required for therapeutic intervention of abdominal Tb complications such as obstruction, perforation, peritonitis, abscess, fistula, or bleeding either before, during, or after initiating anti-Tb therapy [2].

On the basis of the NIS data, perforation was the most common non-diagnostic surgical indication for abdominal Tb in the United States. In contrast, other studies have demonstrated obstruction as the most common reason for surgical intervention, reported in 12%–60% of patients [23–25]. Perforation may be secondary to an obstruction caused by stricture and ulcerations in the dilated proximal loop [23]. It is possible that a proportion of patients who had perforation had a previous pathologic obstruction. Given the breadth of possible presenting symptoms, surgeons operating on patients with an acute abdomen must maintain a high index of suspicion for abdominal Tb. This pathology can masquerade as many, more common diagnoses [2].

The 11% inhospital mortality rate was on the lower range (13%–52%) of mortality rates reported previously from other countries [26,27]. We suspect this lower mortality rate in the NIS dataset may reflect a healthier pre-surgical state among patients treated in the United States, compared with lower resource settings. It is possible persons presenting in the United States may be recognized earlier in the course of their disease as a result of improved diagnostics and access to operation. Timely diagnosis and treatment of abdominal Tb has been shown to impact prognosis significantly [26]. Observed death was as high as 60% for patients in whom treatment was not started within 30 days after initial presentation [26]. Another possible explanation for the lower observed deaths is the increased ability to rescue a patient in extremis with intensive supportive care in higher resource settings.

The risk factors for abdominal Tb—cirrhosis, HIV, malnutrition, immunosuppression, renal failure—are in themselves risk factors for morbidity or death after operation [10,19,27–30]. Among the NIS cohort, we found hypertension and coagulopathy risk factors for death after operation. Notably, this is the first time that these comorbidities have been associated specifically with higher mortality rates in abdominal operation for Tb. The wide CI and high OR make interpretation difficult, however. In addition, it is possible that this observation is an artifact of this being a retrospective analysis. In NIS, there is no distinction made between the temporal relationships of diagnoses; the possibility that coagulopathy and hypertension were pre-existing the diagnoses of an acute abdomen was as likely as diagnoses accrued after development of a surgical abdomen.

This study has several limitations. First, there may be misclassification bias associated with utilizing ICD-9 codes; the integrity of the coding depends on the quality of coders at each participating institution. Second, NIS reports event-level data but does not have unique personal identifiers. An individual who was admitted multiple times over the course of the study period would be counted as multiple cases. Third, ICD-9 coding data does not allow for interpretation of causality, merely co-existence of ICD-9 codes. It is not possible to determine whether Tb was the cause of a patient's acute abdominal pain that resulted in admission, only that they were temporally related during the admission. This may lead to overestimation of the frequency of the acute abdomen in Tb.

Fourth, drug and alcohol use is inconsistently reported in administrative databases, and interpretation of these frequencies must be viewed with caution. Fifth, NIS does not include Indian Health Service or Veterans Administration hospitals, which may lead to an underestimation of the burden of Tb in the United States. Sixth, if a patient with Tb and an acute abdomen were operated on at a hospital reporting to NIS, but transferred to a facility not reporting to NIS where he or she later died, this death would not be captured in NIS. Seventh, the race category typically has a high missing entry rate in NIS. We believe that this will have limited impact in our analysis, however; in our reviewed sample, more than 94% of admissions had an associated race.

Conclusion

Abdominal Tb is an infrequent manifestation of EPTb in the United States, but may cause significant morbidity and death. While anti-Tb microbials are a mainstay of therapy, operation may still be required for diagnosis and treatment of patients with abdominal Tb who present with acute abdominal pain. In the United States, black and Hispanic patients are less likely to receive surgical intervention compared with white or Asian races. The inhospital deaths from acute abdominal pain necessitating operation among patients with Tb are high. An elevated index of suspicion for abdominal Tb in patients with unexplained abdominal symptoms should be maintained given its protean manifestations. Prompt recognition and appropriate treatment of patients with abdominal Tb may reduce the morbidity and deaths observed with this obdurate pathogen.