Risk factors associated with prolonged hospital length-of-stay: 18-year retrospective study of hospitalizations in a tertiary healthcare center in Mexico

Abstract

Background: Hospital length-of-Stay has been traditionally used as a surrogate to evaluate healthcare efficiency, as well as hospital resource utilization. Prolonged Length-of-stay (PLOS) is associated with increased mortality and other poor outcomes. Additionally, these patients represent a significant economic problem on public health systems and their families. We sought to describe and compare characteristics of patients with Normal hospital Length-of-Stay (NLOS) and PLOS to identify sociodemographic and disease-specific factors associated with PLOS in a tertiary care institution that attends adults with complicated diseases from all over Mexico.

Materials and methods: We conducted a retrospective analysis of hospital discharges from January 2000-December 2017 using institutional databases of medical records. We compared NLOS and PLOS using descriptive and inferential statistics. PLOS were defined as those above the 95th percentile of length of hospitalization.

Results: We analyzed 85,904 hospitalizations (1,069,875 bed-days), of which 4,427 (5.1%) were PLOS (247,428 bed-days, 23.1% of total bed-days). Hematological neoplasms were the most common discharge diagnosis and surgery of the small bowel was the most common type of surgery. Younger age, male gender, a lower physician-to-patient ratio, emergency and weekend admissions, surgery, the number of comorbidities, residence outside Mexico City and lower socioeconomic status were associated with PLOS. Bone marrow transplant (OR 18.39 [95% CI 12.50-27.05, p<0.001), complex infectious diseases such as systemic mycoses and parasitoses (OR 4.65 [95% CI 3.40-6.63, p<0.001), and complex abdominal diseases such as intestinal fistula (OR 2.57 [95% CI 1.98-3.32) had the greatest risk for PLOS. Risk of mortality in patients with PLOS increased more than threefold (3.7% vs 13.3%, p<0.001).

Conclusions: We report some key sociodemographic and disease-specific differences in patients with PLOS. These could serve to develop a specific model of directed hospital healthcare for patients identified as in risk of PLOS.

Fig 1. Distribution of prolonged length of stay (PLOS) events by type of hospitalization from 2000–2017.

(A) Overall Distribution of PLOS events by type of hospitalization (elective or emergency and surgical and non-surgical). The frequency of PLOS was significantly higher during hospitalization events that required a surgical intervention. (B) Annual trends of the distribution of type of hospitalization. During the study period there was an important reduction in the proportion of elective and urgent surgical events of hospitalization (dark and light gray bars). Elective hospitalization events for surgical procedures increased the most during the study period.

Fig 2. Adjusted risk of a prolonged length of stay (PLOS) event by diagnosis at hospital discharge.

Odds ratios for PLOS by diagnosis at discharge were adjusted for age, gender, physician-to-patient ratio, type of admission, readmission at 30 days, day of admission (weekday vs weekend), number of additional diagnosis, place of residence and socioeconomic status using multinomial logistic regression models fixing “Diseases of the liver, biliary tract and pancreas (K70.0-K79.9, K83.0-K89.9)” as the reference group.

Fig 3. Annual frequency of hospitalizations classified as prolonged length-of-stay (PLOS) from 2000–2017.

(A) The vertical, gray bars represent the annual percentage of hospitalization events classified as PLOS. The percentage increased from 2.4% in 2000 to 7.6% in 2007, then declined slightly in the ensuing years and remained stable during 2009–2016 with a later peak in 2017. The black, dotted line, summarizes the annual median length-of-stay (LOS) in days across time, during the study period. The median LOS for all hospitalization events was 8 days in 2000, peaked at 10 days in 2006 and 2007 and then declined to 8 days afterwards and up to 2015, when it declined again by one day (B). The black, vertical, boxplots illustrate the annual adjusted odds ratios (aORs) for prolonged stay of hospitalization (PLOS) using 2000 as the year of reference. We used multinomial logistic regression models to control for age, gender, type of admission, recent hospital discharge, weekday/weekend admission, additional diagnoses, place of residence and socioeconomic status, using inverse probability weights based on diagnosis of admission. The adjusted risk of PLOS increased between 2000 and 2007, then substantially and continuously decrease afterwards despite a sustained percentage of PLOS episodes after 2008.