Mortality and its risk factors in Malawian children admitted to hospital with clinical pneumonia, 2001–12: a retrospective observational study

Thursday, 7th of January 2016

Mortality and its risk factors in Malawian children admitted to hospital with clinical pneumonia, 2001–12: a retrospective observational study

Dr Marzia Lazzerini, PhD,

Nadine Seward, PhD,

Norman Lufesi, MPhil,

Rosina Banda, BScNE,

Sophie Sinyeka, MSc,

Gibson Masache, MSc,

Bejoy Nambiar, MPH,

Charles Makwenda, MSc,

Prof Anthony Costello, FMedSci,

Eric D McCollum, MD,

Dr Tim Colbourn, PhD

DOI: http://dx.doi.org/10.1016/S2214-109X(15)00215-6

Creative Commons Attribution – NonCommercial – NoDerivs (CC BY-NC-ND 4.0) How you can reuse

This article can be found in the following collections: Global Health; Paediatric infections; Respiratory tract infections; Paediatric respiratory medicine

Excerpts below; full text is at http://www.thelancet.com/journals/langlo/article/PIIS2214-109X(15)00215-6/fulltext

Summary

Background

Few studies have reported long-term data on mortality rates for children admitted to hospital with pneumonia in Africa. We examined trends in case fatality rates for all-cause clinical pneumonia and its risk factors in Malawian children between 2001 and 2012.

Methods

Individual patient data for children (<5 years) with clinical pneumonia who were admitted to hospitals participating in Malawis Child Lung Health Programme between 2001 and 2012 were recorded prospectively on a standardised medical form. We analysed trends in pneumonia mortality and childrens clinical characteristics, and we estimated the association of risk factors with case fatality for children younger than 2 months, 2–11 months of age, and 12–59 months of age using separate multivariable mixed effects logistic regression models.

Findings

Between November, 2012, and May, 2013, we retrospectively collected all available hard copies of yellow forms from 40 of 41 participating hospitals. We examined 113 154 pneumonia cases, 104 932 (92·7%) of whom had mortality data and 6903 of whom died, and calculated an overall case fatality rate of 6·6% (95% CI 6·4–6·7). The case fatality rate significantly decreased between 2001 (15·2% [13·4–17·1]) and 2012 (4·5% [4·1–4·9]; p_trend<0·0001). Univariable analyses indicated that the decrease in case fatality rate was consistent across most subgroups. In multivariable analyses, the risk factors significantly associated with increased odds of mortality were female sex, young age, very severe pneumonia, clinically suspected Pneumocystis jirovecii infection, moderate or severe underweight, severe acute malnutrition, disease duration of more than 21 days, and referral from a health centre. Increasing year between 2001 and 2012 and increasing age (in months) were associated with reduced odds of mortality. Fast breathing was associated with reduced odds of mortality in children 2–11 months of age. However, case fatality rate in 2012 remained high for children with very severe pneumonia (11·8%), severe undernutrition (15·4%), severe acute malnutrition (34·8%), and symptom duration of more than 21 days (9·0%).

Interpretation

Pneumonia mortality and its risk factors have steadily improved in the past decade in Malawi; however, mortality remains high in specific subgroups. Improvements in hospital care may have reduced case fatality rates though a lack of sufficient data on quality of care indicators and the potential of socioeconomic and other improvements outside the hospital precludes adequate assessment of why case-fatality rates fell. Results from this study emphasise the importance of effective national systems for data collection. Further work combining this with data on trends in the incidence of pneumonia in the community are needed to estimate trends in the overall risk of mortality from pneumonia in children in Malawi.

Funding

Bill & Melinda Gates Foundation.

Introduction

Pneumonia is the leading cause of morbidity and mortality in post-neonatal children under 5 years of age.¹ According to the most recent estimates,¹ 0·9 million children died of pneumonia in 2013, and more than 95% of these deaths happened in low-income and middle-income countries.^1, ²

Few data are available to show the epidemiology and public health burden of paediatric pneumonia cases in African hospitals. A recent systematic review³ identified only 11 studies reporting data on mortality from acute lower respiratory infections in hospitals within the African region; these reports were unpublished, with very few exceptions,⁴ and observation times were limited to 2–3 years.³

Malawi is currently one of the poorest countries in sub-Saharan Africa. However, according to national statistics,⁵ major progress was made in the past 15 years, and Malawi is on track to reach the Millennium Development Goal 4 of a two-thirds reduction in under-5 year mortality from 1990 to 2015.

In 2000, the Malawi Ministry of Health implemented a standardised medical chart for children younger than 5 years who were admitted to hospital with clinical pneumonia. These data have been routinely collected prospectively but never comprehensively analysed. We have analysed the available individual patient data from hospitals in Malawi that implemented this routine system of data collection between 2001 and 2012, with the objective of describing trends in case fatality rates for all-cause clinical pneumonia and its risk factors in children younger than 5 years.

Research in context

Evidence before this study

A recent systematic review identified only 11 studies reporting data on mortality in children admitted to hospital with acute lower respiratory infections in the African region, and with very few exceptions, reports were unpublished, and with an observation time limited to 2–3 years. We also searched PubMed using the following search strategy: (“Pneumonia”[Mesh] OR “Respiratory Tract Infections”[Mesh]) AND (“Child”[Mesh] OR (“child”[MeSH Terms] OR “child”[All Fields] OR “children”[All Fields]) OR (“pediatrics”[MeSH Terms] OR “pediatrics”[All Fields] OR “paediatric”[All Fields]) OR (“pediatrics”[MeSH Terms] OR “pediatrics”[All Fields] OR “pediatric”[All Fields])) AND (“Malawi”[MeSH Terms] OR “Malawi”[All Fields]) from inception to July 8, 2015, with no language restrictions. We found 68 studies, none of which covered the range of years and numbers of hospitals of our study.

Added value of this study

We have analysed an individual patient database of hospitalised cases of pneumonia in children in Malawi, collected over a 12 year period. The data shows a clear decline in case fatality rate between 2001 and 2012, although this rate remains high in some subgroups (children with very severe pneumonia, severe undernutrition, severe acute malnutrition, and symptom duration >21 days).

Implications of all the available evidence

Overall, our study supports the finding from the Millennium Development Goal 4 indicators that under-5 mortality has significantly decreased in Malawi in recent years. Further research is needed to link hospital data with community-based pneumonia data and to investigate quality of care provided to children at different levels of the health system.

Methods

Study design and participants

In 2000, the Malawi Ministry of Healths Acute Respiratory Infection unit (ARI) and the International Union Against Tuberculosis and Lung Disease implemented the Child Lung Health Programme (CLHP),^6, ⁷ which included two key elements: national clinical pneumonia management guidelines⁸adapted from WHO guidelines;^9, ¹⁰ and the implementation of a standardised patient chart (the yellow form) to be used as an official medical file for each child admitted to hospital for pneumonia. CLHP clinical pneumonia was defined according to Malawi ARI guidelines (panel). The yellow form contains individual patient data such as demographic variables, clinical signs and symptoms, pneumonia disease severity, comorbidities, treatments received, and outcomes (appendix pp 3–4). The following criteria were used for CLHP programme participation: an active ARI programme; leadership commitment; one health worker responsible for implementation (local ARI coordinator); and about 100 000 population catchment area.⁶ District government hospitals were prioritised for participation in CLHP, and by 2004, 22 of 23 district hospitals and three of four central government hospitals were enrolled.⁶ In 2005, with the support of the Scottish Government, the programme expanded to include the Christian Hospital Association of Malawi hospitals, which are mostly first-level, fee-based facilities. By 2012, 22 of 23 district hospitals, three of four central hospitals, and 16 of 37 Christian Hospital Association of Malawi facilities were participating in the CLHP (appendix p 5).

Panel

Classification of severity of clinical pneumonia¹⁰

Non-severe pneumonia (2–59 months of age)^**

^*Young infants younger than 2 months do not have a non-severe pneumonia classification.

Cough, difficulty breathing, or both

Fast breathing for age^††

†60 breaths per min or more if child is younger than 2 months; 50 breaths per min or more if child is 2–11 months old; 40 breaths per min or more if child is 12–59 months old.

No lower chest indrawing and no danger signs^‡‡

‡Danger signs are any of the following: central cyanosis, severe respiratory distress (grunting, head nodding, severe chest indrawing), stridor, a general danger sign (inability to drink, breastfeed, or both, lethargy or unconsciousness, convulsions), apnoea (if child is 0–2 months of age). Wheeze is not considered in diagnosis or classification of severity of pneumonia.¹⁰

Severe pneumonia (<2 months of age)

Cough, difficulty breathing, or both, and

Lower chest indrawing or fast breathing for age^**^††

No danger signs^‡‡

Severe pneumonia (2–59 months of age)

Cough, difficulty breathing, or both

Lower chest indrawing

No danger signs^‡‡

Might or might not have fast breathing for age^††

Very severe pneumonia (0–59 months of age)

Cough, difficulty breathing, or both

At least one danger sign^‡‡

Might or might not have fast breathing for age^††

Lower chest indrawing

From 2001 to 2005, major external support was provided to the CLHP: health staff were trained in the programme, which included a follow-up refresher session and on-the-job training; international expert technical guidance that focused on maintaining data quality, accuracy, and completeness was provided twice annually; and programme managers met regularly to review their work.⁶ The CLHP also provided an uninterrupted supply of antibiotic drugs for pneumonia treatment.⁶ Between 2006 and 2008, the programme was gradually transitioned to the Ministry of Health, whereas external assistance focused on enrolling facilities from the Christian Hospital Association of Malawi using previous approaches. In 2009, the Ministry of Health assumed primary control and arranged supervision visits and follow-up trainings according to local needs and available funding.

Between November, 2012, and May, 2013, the study authors and data collectors employed by PACHI (the local research organisation conducting the study) collected all available hard copies of yellow forms from 40 of the 41 participating hospitals, where the forms were stored (usually under the supervision of the local ARI coordinator). Standardised data cleaning and data entry, including systematic quality assurance checks, were done by data entry clerks and nurses under close supervision by ML, RB, and SS. Confidentiality was maintained by de-identifying all files before database entry.

This study was approved by the Ethics Committee of University College London (project number 2006/002) and by the National Health Sciences Research Committee of Malawi (protocol number 941).

Descriptive and statistical analyses

Our descriptive analyses focused on the following outcomes: annual case fatality rates, annual case fatality rates in predefined population subgroups, and changes in childrens characteristics with time. Young infants (defined as infants younger than 2 months) and children age 2–59 months were analysed separately since pneumonia classification and treatment differs in these two groups, according to WHO guidelines.^8, ^9, ¹⁰ Weight-for-age was analysed according to WHO 2006 standards applied retrospectively to the dataset.¹¹ WHO moderate undernutrition and severe undernutrition were defined as weight-for-age ranging between −3 and −2 SD from the median and weight-for-age less than −3 SD from the median, respectively. The clinical diagnosis of severe acute malnutrition was made prospectively during patient care on the basis of visible signs of wasting or oedema of both feet as per Malawi guidelines.^8, ^9, ¹⁰ Pneumonia and fast breathing were defined according to standard WHO criteria (panel). Fever was defined as body temperature greater than 38°C.

We compared mortality rates in patients for whom we had data on basic demographic variables and risk factor variables with rates in patients with missing data using the χ² statistic, Fishers exact test, or t test, as appropriate, to informally investigate the likelihood that missing data could substantially bias subsequent analyses. We also did sensitivity analyses assuming different probabilities of mortality for the cases missing data on mortality relative to cases with data (0, 0·1x, 0·5x, 2x, 10x, and 1, where x is the probability of mortality in cases with data).

We compared rates by year using the χ² test for trend (Cochrane–Armitage test). Univariable logistic regression was used to calculate odds ratios (ORs) for mortality associated with each risk factor in each age group and in different population subgroups; survival time until death was not recorded (appendix p 3–4), precluding Cox regression.

For all statistical tests, a p value of 0·05 or less (two-sided) was considered significant. We used Stata version 13·1 for all analyses.¹²

Three multivariable logistic regression models were created to determine risk factors for mortality for pneumonia cases admitted to hospital in children younger than 2 months, age 2–11 months, and age 12–59 months separately. We used bootstrap resampling of stepwise backwards elimination with a p value cutoff greater than 0·05 for leaving the model to build multivariable models for each age group.^13, ¹⁴ All variables except HIV and malaria (risk factors for which too much data were missing) were initially entered into each of 100 bootstrap replications, and variables that were selected in more than 50% of the replications were included in the final models.^13, ¹⁴ As pneumonia cases were included from 40 hospitals, mortality could be correlated within hospitals. Likelihood ratio test confirmed the clustered nature of the data, and we used mixed effects logistic regression models to account for this. Results of the multivariable logistic regression analyses are presented as adjusted OR (AOR) with 95% CI. Events per variable¹⁵ and area under the receiver-operator-characteristic (ROC) curve are presented as measures of model adequacy and goodness of fit.

To reduce bias and loss of information due to missing data, we used multiple imputation with the assumption that data was missing at random. Data showed evidence of clustering, so we used REALCOM impute software to impute missing data.¹⁶ Variables in the multiple imputation models included the mortality outcome, risk factors of mortality included in the multivariable models, and key variables that were found to be predictors of missingness including region and type of facility. The HIV and malaria variables were excluded for having too many missing data to be imputed. Once we imputed the data in REALCOM, the data were uploaded into Stata for analysis using the commands “mi estimate: xtmelogit”.¹²

We used separate logistic regression models to analyse the available data for HIV. Scenarios were generated to assess the effect of possible trends of HIV prevalence and case fatality rates in HIV-negative cases on overall case fatality rates, assuming odds of mortality in HIV-positive pneumonia cases 5·9 times higher than the odds of mortality in HIV-negative cases¹⁷ in five scenarios and changing the odds in another scenario.

Role of the funding source

The funder of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

All clinical pneumonia cases registered on the yellow forms with an age younger than 5 years were included (figure 1). Of the 113 154 unique cases within the dataset that were available for analysis (figure 1), 102 294 cases were aged 2–59 months and 10 860 cases were younger than 2 months. The number of cases with each potential risk factor in young infants and children who survived, died, or for whom data on case fatality were missing are listed in table 1. Overall, less than 20% of data were missing for all but three variables: previous antibiotic treatment (25·1% of data were missing), malaria blood film (74·8% of data were missing), and HIV status (92·3% of data were missing). The number of annual cases can be found in the appendix (p 6)). District hospitals contributed the largest number of cases (appendix p 7).