The burden of multimorbidity-associated acute hospital admissions in Malawi and Tanzania: a prospective multicentre cohort study

Abstract

Background: The global burden of multimorbidity-the coexistence of two or more long-term conditions-is increasing. Limited access to primary care in sub-Saharan Africa means acute hospital admission is often the sentinel multimorbidity presentation. This prospective multicentre cohort study aimed to describe the burden, constituent diseases, and outcomes of multimorbidity among patients acutely admitted to hospital in Malawi and Tanzania.

Methods: Adults (ie, those aged ≥18 years) admitted to four hospitals (two tertiary and two district hospitals) with acute medical conditions were consecutively recruited within 24 h of presentation and followed up for 90 days. We estimated the prevalence of HIV infection, diabetes, hypertension, and chronic kidney disease using commercially available point-of-care tests, and captured self-reported and clinical diagnoses (n/N [%]). Health economic data were summarised by median and IQR and modelled using generalised linear models. All-cause 90-day mortality was summarised with Kalplan-Meier plots and analysed using Cox regression models.

Findings: 1407 adults (657 [46·7%] were female and 750 [53·3%] were male; mean age was 52·3 years [SD 18·4]) were recruited. We examined multimorbidity prevalence in 1007 participants admitted to three hospitals that accept admissions directly from the community. Multimorbidity was found in 473 (47·0%) of 1007 participants and 292 (29·0%) had a single long-term condition. Outcomes at 90 days were determined for 1317 (93·6%) of 1407 participants. Adjusted 90-day mortality was higher in participants with multimorbidity (335 [41·7%] of 804; hazard ratio 1·5 [95% CI 1·1-2·1]) and those with one long-term condition (80 [28·3%] of 283; 1·5 [1·0-2·1]); compared with those with no long-term conditions (31 [13·5%] of 230). Health-related quality of life was lower in participants with multimorbidity compared with those with one long-term condition (median 0·402 [IQR -0·037 to 0·644] vs 0·557 [0·140 to 0·730]; p=0·005) at baseline, and at final observation (0·858 [0·667 to 1·00] vs 1·00 [0·589 to 1·00] respectively; p=0·01). In Tanzania, medical costs incurred by patients were higher in participants with multimorbidity compared with those with one long-term condition (relative effect 5·77 [95% CI 2·99-11·15]; p<0·0001).

Interpretation: Multimorbidity is common in patients admitted to hospital in Malawi and Tanzania and associated with worse survival and increased cost. Multimorbidity is an urgent public health threat that requires fundamental health-care delivery reform to address population needs.

Funding: National Institute for Health and Care Research and Wellcome Trust.

Translations: For the Chichewa and Kiswahili translations of the abstract see Supplementary Materials section.

Figure 1

Study profile

Figure 2

Prevalence and characteristics of multimorbidity among people admitted to hospital in Malawi and Tanzania This figure summarises disease prevalence across two hospitals in Malawi (ie, the Queen Elizabeth Central Hospital and the Chiradzulu District Hospital) and one in Tanzania (ie, the Hai District Hospital). The x-axis shows the conditions included. Numbers represent the count (n) and percentage (%). When single circles are shown, the corresponding vertical barchart shows the prevalence of the single condition. When there are two or more circles, the corresponding vertical barchart shows the prevalence of coexisting conditions. For example, above the green filled circle for HIV alone, the vertical bar shows the prevalence of participants with HIV alone (112 [11·1%] participants); above the orange filled circles for both HIV and diabetes, the vertical barchart shows the prevalence of participants with coexistent HIV and diabetes but no additional conditions (33 [3·3%] participants). The horizontal barchart shows the prevalence of participants with each individual condition, stratified by colour to depict the number of additional coexistent conditions. Primary conditions diagnosed through study procedures (ie, HIV, hypertension, diabetes, chronic kidney disease, and depression) are highlighted light grey; secondary conditions captured from clinical records (ie, heart failure, cerebrovascular accident, chronic liver disease, chronic obstructive pulmonary disease, and ischaemic heart disease) are highlighted in darker grey. The dark green bar represents those with no long-term conditions.

Figure 3

Kaplan–Meier survival plots, by number of long-term conditions (A) Unadjusted Kaplan–Meier plot. (B) Confounder-adjusted survival curves, based on the Cox regression model with adjustments for age, sex, universal vital assessment score, and site. (C) Box and whisker plot of HRQoL utility scores at baseline, and among survivors at the final observation. p values reflect multivariable generalised linear model analyses (gamma distribution), adjusted by age, sex, site, and universal vital assessment (baseline); and age, sex, site, and number of days in study (final observation). No long-term conditions, n=242; single long-term condition, n=300; and ≥2 long-term conditions, n=865. HRQoL=health-related quality of life.