Prevalence and predictors of prolonged length of stay among patients admitted under general internal medicine in a tertiary government hospital in Manila, Philippines: a retrospective cross-sectional study

We performed a cross-sectional study through retrospective review of medical records of patients admitted at the emergency department, general medicine wards, and the medical intensive care unit of the Philippine General Hospital. The study protocol was approved by the University of the Philippines Manila Review Ethics Board (UPMREB 2020–506-01).

We conducted the study in the University of the Philippines—Philippine General Hospital (UP-PGH), a public teaching hospital and a national tertiary referral center in Manila, Philippines with a 1500-bed capacity [8].

The Department of Medicine is the largest clinical department in UP-PGH, with 13 subspecialty divisions. It includes both outpatient and inpatient services, either as service (i.e. fully subsidized) patients or private (Pay) patients. Service patients are admitted either directly to the wards or Medical Intensive Care Unit (MICU) or at the emergency department then transferred to the wards or the MICU. All admitted service patients are managed by the primary attending service, composed of the service consultant, senior resident-in-charge, junior resident-in-charge, and co-managing fellows-in-training. The department also receives referrals from other departments for co-management and preoperative evaluation which are assigned to a general internal medicine service.

We included all hospital admissions to the General Internal Medicine service between January 1, 2019 to December 31, 2019 that fulfill the following inclusion criteria: age 19 years and older, admitted at least 24 h under the General Internal Medicine as a service patient regardless of area, and General Internal Medicine as the primary attending service on admission. The following patients were excluded: those under the primary service of other departments, those transferred to or from another service or department, or those transferred to or from the Pay services. For patients with multiple admissions, we considered each admission separately.

We classified eligible hospital admissions as either elective or emergency, according to the acuity of the reason for admission. Elective admissions were those directly admitted to the general medicine wards or MICU for non-urgent, elective procedures such as percutaneous coronary intervention, imaging-guided biopsy, and blood transfusions. Admissions through the emergency department for acute urgent or emergent problems that were eventually discharged directly from the ER or transferred to the wards or the MICU were considered emergency admissions.

Length of stay was defined as the time from the day of admission at the emergency room or wards to the last day of hospitalization. As per hospital policy, prolonged length of stay (PLOS) was defined as 14 days or longer for emergency admissions and 3 days or longer for elective admissions; otherwise, it was considered as normal length of stay (NLOS). We obtained the following variables on admission: age, sex, distance of place of residence from the hospital, highest educational attainment, employment status, Medical Social Service classification, smoking status, level of alcohol consumption, functional status, comorbidities, Charlson Comorbidity Index score, and history of prior hospitalization in the past 30 days. We also reviewed the records for the following variables: type of admission (emergency or elective), day of admission, time of admission, number of medications on admission, need for intravenous antibiotics and duration, duration of emergency room and intensive care unit stay, need for invasive and non-invasive ventilation and duration, performance of procedure and surgery, type of surgery and surgical risk of non-cardiac surgeries, need for blood transfusion, need for dialysis, development of shock, type, and duration, development of in-hospital complications and healthcare-associated infections, number of co-managing services, presence of signed advance directive, outcome of hospitalization, and cause of death, if applicable. Their corresponding operational definitions are detailed in Supplementary Table 1. Direct medical costs based on hospital bills, which excluded professional fees of health personnel, were also determined.

Reasons for delay in discharge were reviewed in the weekly census of overstaying patients. Two independent adjudicators classified them as administrative (e.g. delay in procedure schedules, lack of blood products), disease-related (e.g. completion of intravenous antibiotics, difficulty in weaning, need for workup, development of new medical problems, need for palliative care), or patient-related (e.g. home care issues, caregiver issues, financial issues). A third adjudicator was called in cases where the two independent reviewers had conflicting classifications.

Sample size was computed to be 344 using G*Power 3.1 with a 95% confidence level and a power of 0.8 using the odds ratio on risk factors for prolonged hospital stay [4, 9]. We used a simple random sampling method. All admissions that fulfilled the inclusion criteria were encoded in Microsoft Excel and assigned a random number through its random number generator function. The list was sorted in ascending order according to the random numbers generated and served as the study’s sampling frame. Eligible admissions were enrolled consecutively until the desired sample size was reached.

Categorical data were expressed as frequencies and percentages while continuous variables were summarized using median and interquartile range. The median, interquartile range, minimum, and maximum of length of stay and direct medical costs and prevalence of PLOS were calculated. Characteristics of admissions of PLOS and NLOS were compared using t-test, chi-square test, Fisher’s exact test, and Mann–Whitney U test.

Multiple logistic regression analysis was used to evaluate the association of marginally associated variables on crude logistic regression analysis with p-values of at least < 0.25 and having frequencies of at least 5 on all cells. No imputation was done for missing data. The effect sizes from the multivariate analyses are reported as odds ratio. Confidence interval was set at 95%, and a p-value less than 0.05 was considered significant. STATA 16 was used for the analyses [10].