Search tips
Search criteria 


Logo of jgimedspringer.comThis journalToc AlertsSubmit OnlineOpen Choice
J Gen Intern Med. Apr 2003; 18(4): 242–247.
PMCID: PMC1494845
Predictors of Early Hospital Readmission in HIV-infected Patients with Pneumonia
Anita Palepu, MD, MPH,1,2 Huiying Sun, MSc, PhD,2,4 Laura Kuyper, BA,2 Martin T Schechter, OBC, MD, PhD,2,3,4,5 Michael V O'Shaughnessy, OBC, PhD,2,3,4,5,6 and Aslam H Anis, PhD2,3,4,5
1Received from the Department of Internal Medicine, St. Paul's Hospital
2The Centre for Health Evaluation and Outcome Sciences, St. Paul's Hospital
3The British Columbia Centre for Excellence in HIV/AIDS, St. Paul's Hospital
4The Canadian HIV Trials Network, Vancouver, British Columbia, Canada.
5The Department of Health Care and Epidemiology, Vancouver, British Columbia, Canada.
6The Department of Pathology, University of British Columbia, Vancouver, British Columbia, Canada.
Address correspondence and requests for reprints to Dr. Anis: BC Centre for Excellence for HIV/AIDS, St. Paul's Hospital, 570-1081 Burrard St., Vancouver, BC, Canada V6Z 1Y6 (e-mail: anis/at/
Although hospitalization patterns have been studied, little is known about hospital readmission among HIV-infected patients in the era of highly active antiretroviral therapy. We explored the risk factors for early readmission to a tertiary care inner-city hospital among HIV-infected patients with pneumonia in Vancouver, Canada.
Case-control study.
Tertiary care, university-affiliated, inner-city hospital.
All HIV-infected patients who were hospitalized with Pneumocystis carinii pneumonia (PCP) or bacterial pneumonia (BP) between January 1997 and December 2000. Case patients included those who had early readmissions, defined as being readmitted within 2 weeks of discharge (N = 131). Control patients were randomly selected HIV-infected patients admitted during the study period who were not readmitted within 2 weeks of discharge (N = 131), matched to the cases by proportion of PCP to BP.
Sociodemographic, HIV risk category, and clinical data were compared using χ2 test for categorical variables, and the Wilcoxon rank-sum test was used for continuous variables. Multivariable logistic regression was performed to determine the factors independently associated with early readmission. We also reviewed the medical records of 132 patients admitted to the HIV/AIDS ward during the study period and collected more detailed clinical data for a subanalysis.
Patients were at significantly increased odds of early readmission if they left the hospital against medical advice (AMA) (adjusted odds ratio [OR], 4.26; 95% confidence interval [95% CI], 2.13 to 8.55), lived in the poorest urban neighborhood (OR, 2.03; 95% CI, 1.09 to 3.77), were hospitalized in summer season (May though October, OR, 2.36; 95% CI, 1.36 to 4.10), or had been admitted in the preceding 6 months (OR, 2.55; 95% CI, 1.46 to 4.47). Gender, age, history of AIDS-defining illness, and injection drug use status were not significantly associated with early readmission.
Predictors of early readmission of HIV-infected patients with pneumonia included: leaving hospital AMA, living in the poorest urban neighborhood, being hospitalized in the preceding 6 months and during the summer months. Interventions involving social work may address some of the underlying reasons why these patients leave hospital AMA and should be further studied.
Keywords: case-control, hospital readmission, HIV, AIDS, bacterial pneumonia, PCP, antiretroviral therapy
Many studies have indicated that hospital readmissions occur frequently, especially among patients with AIDS, renal disease, cancer, and congestive heart failure.13 Rates of readmissions have ranged from 6.5% within 14 days of discharge among HIV-infected patients with bacterial pneumonia4 to 23% within 12 months of discharge among lower-income patients with heart failure.5 Repeated admissions for chronic medical problems are actually more costly for total days of stay than single, cost-intensive stays.6 Factors shown to influence the likelihood of readmission include socioeconomic status,5,7 receipt of welfare,1 having AIDS,2 prior admissions,3,4,8 use of crack cocaine, leaving hospital unaccompanied by family or friend, and patient noncompliance.1,4
Grant et al. found that AIDS patients with pneumonia who left hospital alone, used crack cocaine, who had 1 or more coincident AIDS diagnosis, and who had been hospitalized in the preceding 6 months were more likely to be readmitted within 2 weeks of discharge.4 However, this study was done prior to the widespread availability of highly active antiretroviral therapy (HAART), i.e., combination therapy with at least 3 antiretroviral agents. Since it is well known that HAART provided a therapeutic breakthrough that led to significant reductions in morbidity and mortality as well as enhanced quality of life for many HIV-positive individuals,911 its impact on hospital readmission remains to be investigated.
Hospital admission for acute illness among HIV-infected patients may serve as an opportunity to review patients' medications and to optimize therapy prior to discharge. The uptake of HAART may be an important marker of quality of care among HIV-positive patients. We therefore explored the factors associated with early readmission to a tertiary care inner-city hospital among HIV-positive patients with pneumonia as well as the uptake of HAART at admission and discharge among a subgroup of these patients in Vancouver, Canada.
All admissions records, from January 1, 1997 to December 31, 2000, of HIV-infected patients with Pneumocystis carinii pneumonia (PCP) (International Classification of Diseases, Ninth Revision [ICD-9] code 136.3) or bacterial pneumonia (BP) (ICD-9 codes 481–483, 485, and 486) were retrieved from discharge abstracts. A discharge abstract, completed by the staff of the Medical Records Department, is a patient-specific administrative record from the start of admission to discharge or transfer to another institution or death. Patient demographics such as age, gender, and postal code, and clinical information on diagnosis, procedures, and patients' injection drug use were recorded in the discharge abstract. Case patients were defined as all HIV-infected patients with a diagnosis of PCP or BP for at least 1 hospitalization who were subsequently readmitted to St. Paul's Hospital within 14 days for any reason (early readmission). If a case patient had more than 1 admission for BP or PCP and was subsequently readmitted within 14 days during the study period, then 1 of these admissions was randomly selected for the study. The selected admission was defined as the index hospitalization of case patients. Control patients were randomly selected from patients who had hospitalizations for BP or PCP and who were not readmitted within 14 days. In patients hospitalized more than once for BP or PCP during the study period, only 1 (randomly selected) hospitalization was included and was defined as the index hospitalization for the controls. Case and control patients were matched by proportion of PCP and BP.
Based on the information from the index hospitalization, patients were classified as injection drug users or non–injection drug users. Patients with at least 1 diagnosis of AIDS or advanced HIV disease (ICD-9 codes 42.0–44.9) were identified. This classification was used as a marker of the HIV disease severity. According to patients' postal codes, they were classified as “Downtown Eastside” (the poorest urban neighborhood in Canada and subsequently referred to as the “poorest urban neighborhood”),12“no fixed address” if they did not report a fixed address at admission, and “other.” We classified patients as leaving the hospital against medical advice (AMA) and not-AMA based on the manner in which they were discharged from the index hospitalization. Patients were also classified as having a previous admission if they were admitted in the 6 months preceding their index hospitalizations. To look at the effect of season on risk of readmission of HIV-positive patients with pneumonia, we classified the index hospitalization into 2 seasons, summer (May through October) and winter (November through April), hypothesizing that most of the admissions would be in the winter months.
To obtain more detailed clinical data, including ethnicity (Aboriginal versus other) and uptake of HAART at admission and discharge, we reviewed the medical records of 368 patients admitted with PCP or BP to the St. Paul's Hospital HIV/AIDS ward during the study period. We employed the same procedures as described earlier to identify case and control patients. Finally, we compared the coding of injection drug use in the hospital data set to the 30-day drug use history obtained from 132 patients in the HIV/AIDS ward and found the κ statistic for reliability was 0.68 (95% confidence interval [95% CI], 0.55 to 0.81). The St. Paul's Hospital Ethics Committee for Human Experimentation approved the data collection for this study.
Both bivariable and multivariable analyses were used in the study. In the bivariable analysis, categorical variables were compared using χ2 test, while continuous variables were compared using the Wilcoxon rank-sum test. Variables were checked for co-linearity prior to applying logistic regression to determine the factors associated with early readmissions. We also checked the effect of possible interaction terms, such as AMA and area of residence, AMA and injection drug use (IDU), and IDU and area of residence. The same procedure was repeated for the subanalysis using the HIV/AIDS ward data. All statistical analyses were performed using the SAS 8.2 software program (SAS Institute, Inc., Cary, NC).
Between January 1, 1997 and December 31, 2000, 724 HIV-infected patients accounted for 1,311 admissions to St. Paul's Hospital for PCP or BP. Of the 287 patients who were hospitalized for PCP, 60 (21%) had a subsequent readmission within 14 days; and of the 1,024 BP cases, 156 (15%) had a subsequent early readmission. Of the 724 patients admitted during the study time period, 140 (19%) incurred at least 1 early readmission. The median time to readmission for the case patients was 5 days (interquartile range [IQR], 2–9).
Case and control patients were matched by proportion of PCP and BP. There were 27 PCP and 104 BP case patients matched to 27 PCP and104 BP control patients. Table 1 presents patients' characteristics. We found that case patients were more likely to be injection drug users, live in the poorest urban neighborhood, leave the hospital AMA, be hospitalized in summer season, and have had an admission to hospital in the preceding 6 months. Length of stay (LOS) at the index hospitalization did not differ significantly between case and control patients (median LOS 7, days versus 6 days; P = .14). The median CD4 cell count for cases was lower than for the controls (65 versus 160 cells/mm3; P = .009).
Table 1
Table 1
Characteristics of Case and Control Patients at the Index Hospitalization
Table 2 lists the unadjusted and adjusted odds ratios (ORs) and corresponding 95% confidence intervals (95% CIs) of factors associated with early readmission. After adjusting for gender, history of an AIDS-defining illness, and IDU status, case patients were more likely to leave the hospital AMA (OR, 4.26; 95% CI, 2.13 to 8.55), to have a previous admission <6 months (OR, 2.55; 95% CI, 1.46 to 4.47), to live in the poorest urban neighborhood (OR, 2.03; 95% CI, 1.09 to 3.77), and to be hospitalized during the summer season (OR, 2.36; 95% CI, 1.36 to 4.10). Since none of the interaction terms were statistically significant, we did include them in the final logistic regression model.
Table 2
Table 2
Factors Associated With Early Hospital Readmission: Unadjusted and Adjusted Odds Ratios Among HIV Patients Hospitalized With BP or PCP During Their Index Hospitalization
For the subanalysis, there were 368 HIV-infected patients who incurred 595 admissions to the HIV/AIDS ward at St. Paul's Hospital for PCP or BP between January 1, 1997 and December 31, 2000. Of the 152 hospital admissions for PCP, 31 (20%) had a subsequent readmission within 14 days; and of the 443 BP cases, 70 (16%) had a subsequent early readmission. Of the 368 patients admitted during the study time period, 69 (19%) required at least 1 early readmission. There were 19 PCP and 47 BP case patients matched to 19 PCP and 47 BP control patients. The median time to readmission for the case patients was 6 days (IQR, 3–10).
Table 3 depicts the characteristics of cases and controls who had admissions to the HIV/AIDS ward. Cases were more likely to be female, Aboriginal, live in the poorest urban neighborhood, be hospitalized during the summer months, leave hospital AMA, and have an admission in the previous 6 months. In terms of receipt of HAART, there was no significant difference between cases and controls in the proportion on HAART at admission (26% vs 14%; P = .08) and although a similar proportion were discharged on HAART, more controls were scheduled to see a prescribing physician to start HAART in the community (26% vs 9%; P = .03). The median CD4 cell counts for cases and controls were similar (160 vs 90 cells/mm3; P = .08). We also found that patients who left AMA were much more likely not to have HAART prescribed compared to those who were discharged formally (Table 4). Table 5 shows the unadjusted and adjusted odds ratios of factors associated with early hospital readmission. Similar to the main analysis, living in the poorest urban neighborhood, leaving hospital AMA, and having a hospital admission in the previous 6 months increased the odds of readmission.
Table 3
Table 3
Characteristics of Case and Control Patients from the HIV/AIDS Ward at Their Index Hospitalization
Table 4
Table 4
Receipt of HAART at Discharge for Case and Control Patients from the HIV/AIDS Ward Stratified by AMA Status
Table 5
Table 5
Factors Associated With Early Hospital Readmission: Unadjusted and Adjusted Odds Ratios Among HIV Patients Hospitalized With BP and PCP During Their Index Hospitalization on the HIV/AIDS Ward
Among HIV-positive patients who were hospitalized with PCP and bacterial pneumonia, we found the following risk factors for early readmission: leaving hospital AMA, living in the poorest neighborhood, the index admission occurring during the summer months and being hospitalized in the preceding 6 months. Our 2-week readmission rates were substantially higher than those reported by Grant et al.4 (PCP: 21% vs 4.2%; BP: 15% vs 6.5%). Our patients had higher median CD4 cell counts than Grant's cohort (cases, 65 vs 19 cells/mm3; controls, 160 vs 26 cells/mm3), who appeared to be in more-advanced stages of AIDS. This large difference in readmission rates may be due to differences in population and health care systems. Furthermore, the difference in CD4 counts between cases and controls may reflect the possibilities that patients with readmissions were more likely to have had HIV for longer periods of time, that they had been less likely to take HAART, or that they had higher rates of nonadherence or drug therapy resistance or failure.
Leaving hospital against medical advice was the strongest predictor of readmission. In an earlier study, we found that leaving on welfare check day and a history of IDU were significant predictors of discharge AMA among hospitalized HIV-infected patients.13 In addition, these patients leaving AMA, when compared to those formally discharged, were found to be readmitted more frequently (frequency ratio:95% C.I., 1.25:1.11 to 1.42), be more likely to be readmitted with a related diagnosis within 30 days (OR, 5.00; 95%CI, 3.04 to 8.24) and to have significantly longer length of stay in the follow-up period. Other studies have found that patients who leave AMA tend to be younger,14,15 to be of a lower income level,15 and to have no primary care physician.14 As in our earlier study,13 patients who leave AMA are more likely to be substance users,14,16 and to use psychoactive drugs more frequently than patients discharged formally.16 In addition, patients who are discharged AMA may have negative preconceptions about hospital stays and feel that hospital staff are not helpful.17,18 Because leaving hospital AMA is an indicator of incomplete therapy, it may negatively affect the health outcomes of the patients,14 leading to rapid rehospitalization.16,17 Patients who leave AMA are up to 7 times more likely to be readmitted than those formally discharged.14 In addition, Weingart et al. suggest that a patient who leaves AMA may not be able to remain in hospital for the duration of treatment for personal reasons, and would return after they have addressed other needs.15 Low income is correlated with AMA discharges,15 and among HIV-infected persons residing in the poorest neighborhood, welfare check issue day may influence their decision to leave the hospital prematurely.13
Grant et al. found that leaving the hospital unaccompanied by family or friend was independently associated with hospital readmission.4 This factor may represent a sensitive measure of social isolation even better than living situation, because living situation (as well as other indicators of social isolation) was not different between cases and controls. Illicit drug use and poor social support also have been associated with higher rehospitalization rates.1,4,1921 Another study found that schizophrenic patients whose family members came to visit them were less likely to have early rehospitalizations.19 They suggested that the social environment of the discharged individual might influence the course of recovery. The problems associated with lack of social support are important to examine in the context of our study population. In our cohort, we found that readmissions were significantly associated with living in the poorest urban neighborhood, where illicit drug use is alarmingly high.22
As Grant et al.4 hypothesized, concomitant AIDS diagnoses were not predictive of early hospital readmission in the era of HAART; however, recent antecedent hospitalization continues to be associated with readmission and may reflect suboptimal management of HIV disease. There was a trend for the controls to be discharged on HAART or scheduled to see a prescribing physician regarding HAART as an outpatient compared to cases. Unfortunately, more case patients left hospital AMA, precluding arrangement for this complex therapeutic regimen.
Prior admission within the past 6 months was associated with early readmission, which has been confirmed elsewhere.4 The number of prior admissions has been found to be a strong predictor of future rehospitalization.23 One study showed that patients who were readmitted were significantly more likely than those who were not readmitted to have had 4 or more prior hospitalizations.19 This may be explained by the notion that frequently hospitalized patients tend to be chronically ill and are often rehospitalized for the same illness.20 In addition, we unexpectedly found that hospitalization in the summer months was associated with higher risk for readmission. The reason for this is unclear, because we controlled for leaving AMA that may occur more frequently in the summer months.
Limitations of our study include retrospective data collection and clinical data that were limited to the subsample of patients on the HIV/AIDS ward. As in the study by Grant et al.,4 during the pre-HAART era, we limited our diagnoses to BP and PCP, because they are common and reliably diagnosed. Data on non-injection drug use were not specifically collected. We probably underestimated the number of readmissions, because we studied only 1 hospital, although the patients served by this hospital tend to come back for their medical care because it is a well-known AIDS tertiary care center with a specialized HIV/AIDS ward.24,25 It is possible, however, that injection drug users are being readmitted elsewhere and this may explain why it was not significantly associated with hospital readmission. In addition, the patients from the poorest urban neighborhood who leave AMA may be the type of patient who would re-present to the same hospital. The sample size may have been too small in the subanalysis to identify other factors of interest that are associated with readmission. Finally, the odds ratios derived from case-control are somewhat inflated, and a longitudinal study would provide more precise estimates.
On the basis of our findings, we recommend identifying HIV-infected patients who are at risk for early readmission, such as those with recent hospitalizations and those who live in poorer urban neighborhoods. A social worker might be able to explore and address some of the underlying reasons that these patients leave hospital AMA; this could alleviate the health and economic burden of potentially avoidable readmissions in this vulnerable population.
Dr. Palepu holds a Canadian Institute for Health Research New Investigator Award. Dr. Schechter is a recipient of a Canada Research Chair.
1. Frankl SE, Breeling JL, Goldman L. Preventability of emergent hospital readmission. Am J Med. 1991;90:667–74. [PubMed]
2. Phillips RS, Safran C, Cleary PD, Delbanco TL. Predicting emergency readmissions for patients discharged from the medical service of a teaching hospital. J Gen Intern Med. 1987;2:400–5. [PubMed]
3. Weissman JS, Ayanian JZ, Chasan-Taber S, Sherwood MJ, Roth C, Epstein AM. Hospital readmissions and quality of care. Med Care. 1999;37:490–501. [PubMed]
4. Grant RW, Charlebois ED, Wachter RM. Risk factors for early hospital readmission in patients with AIDS and pneumonia. J Gen Intern Med. 1999;14:531–6. [PMC free article] [PubMed]
5. Philbin EF, Dec GW, Jenkins PL, DiSalvo TG. Socioeconomic status as an independent risk factor for hospital readmission for heart failure. Am J Cardiol. 2001;87:1367–71. [PubMed]
6. Anderson G, Steinberg E. Hospital readmission in the Medicare population. N Engl J Med. 1984;311:1349–53. [PubMed]
7. Glazier RH, Badley EM, Gilbert JE, Rothman L. The nature of increased hospital use in poor neighbourhoods: findings from a Canadian inner city. Can J Public Health. 2000;91:268–73. [PubMed]
8. Weber AE, Yip B, O'Shaughnessy MV, Montaner JS, Hogg RS. Determinants of hospital admission among HIV-positive people in British Columbia. CMAJ. 2000;162:783–6. [PMC free article] [PubMed]
9. Hogg RS, Yip B, Kully C, et al. Improved survival among HIV-infected patients after initiation of triple-drug antiretroviral regimens. CMAJ. 1999;160:659–65. [PMC free article] [PubMed]
10. Revicki DA, Moyle G, Stellbrink HJ, Barker C. Quality of life outcomes of combination zalcitabine-zidovudine, saquinavir-zidovudine, and saquinavir-zalcitabine-zidovudine therapy for HIV-infected adults with CD4 cell counts between 50 and 350 per cubic millimeter. PISCES (SV14604) Study Group. AIDS. 1999;13:851–8. [PubMed]
11. Palella FJ, Jr, Delaney KM, Moorman AC, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators. New Engl J Med. 1998;338:853–60. [PubMed]
12. Statistics Canada. 1996 Population Census of Canada. Ottawa: Statistics Canada; 1997.
13. Anis AH, Sun H, Guh DP, Palepu A, Schechter MT, O'Shaughnessy MV. Leaving hospital against medical advice among HIV-positive patients. CMAJ. 2002;167:633–7. [PMC free article] [PubMed]
14. Jeremiah J, O'Sullivan P, Stein MD. Who leaves against medical advice? J Gen Intern Med. 1995;10:403–5. [PubMed]
15. Weingart SN, Davis RB, Phillips RS. Patients discharged against medical advice from a general medicine service. J Gen Intern Med. 1998;13:568–71. [PMC free article] [PubMed]
16. Pages KP, Russo JE, Wingerson DK, Ries RK, Roy-Byrne PP, Cowley DS. Predictors and outcome of discharge against medical advice from the psychiatric units of a general hospital. Psychiatr Serv. 1998;49:1187–92. [PubMed]
17. Dalrymple AJ, Fata M. Cross-validating factors associated with discharges against medical advice. Can J Psychiatry. 1993;38:285–9. [PubMed]
18. Targum SD, Capodanno AE, Hoffman HA, Foudraine C. An intervention to reduce the rate of hospital discharges against medical advice. Am J Psychiatry. 1982;139:657–9. [PubMed]
19. Olfson M, Mechanic D, Boyer CA, Hansell S, Walkup J, Weiden PJ. Assessing clinical predictions of early rehospitalization in schizophrenia. J Nerv Ment Dis. 1999;187:721–9. [PubMed]
20. Haywood TW, Kravitz HM, Grossman LS, Cavanaugh JL, Jr, Davis JM, Lewis DA. Predicting the “revolving door” phenomenon among patients with schizophrenic, schizoaffective, and affective disorders. Am J Psychiatry. 1995;152:856–61. [PubMed]
21. Havassy BE, Arns PG. Relationship of cocaine and other substance dependence to well-being of high-risk psychiatric patients. Psychiatr Serv. 1998;49:935–40. [PubMed]
22. Strathdee SA, Patrick DM, Currie SL, et al. Needle exchange is not enough: lessons from the Vancouver injecting drug use study. AIDS. 1997;11:F59–65. [PubMed]
23. Korkeila JA, Lehtinen V, Tuori T, Helenius H. Frequently hospitalised psychiatric patients: a study of predictive factors. Soc Psychiatry Psychiatr Epidemiol. 1998;33:528–34. [PubMed]
24. Palepu A, Strathdee S, Hogg R, et al. The social determinants of emergency department and hospital use by injection drug users in Canada. Urban Health. 1999;76:409–18. [PMC free article] [PubMed]
25. Palepu A, Tyndall MW, Leon H, et al. Hospital utilization and costs in a cohort of injection drug users. CMAJ. 2001;165:415–20. [PMC free article] [PubMed]
Articles from Journal of General Internal Medicine are provided here courtesy of
Society of General Internal Medicine