An evaluation of inpatient morbidity and critical care provision in Zambia

Introduction

The recent Lancet Commission on Global Surgery highlighted surgery as a cost-effective, indivisible and indispensable part of health care globally 1. Currently, five billion people lack access to safe, timely, affordable surgery and anaesthetic care when needed. Critical care is a vital part of safe surgical and appropriate postoperative care, and as part of a wider investment in surgical services, it has been shown to be affordable, save lives and promote economic growth 2.

Although comprehensive data on the number of critical care beds in sub-Saharan Africa are not available, previous research describing the ‘supply’ side of critical care provision indicated a likely shortfall. A continent-wide survey of 263 anaesthetic practitioners was carried out at the All-African Anaesthesia Congress in 2009 to determine the availability of critical care services, and the ability to implement the Surviving Sepsis Campaign (SSC) care bundle 3-5. This study found that 25% of anaesthesia practitioners in low- and lower middle-income African countries had no critical care provision at their place of work. Furthermore, only 1.2% of respondents had the facilities to implement the full Surviving Sepsis Campaign care bundle. Although this result must be interpreted with caution (the full care bundle at that time included activated protein C for example), many more integral interventions were commonly unavailable. Noradrenaline, invasive arterial blood pressure monitoring and mechanical ventilation were either unavailable or not always available in 53.3%, 76.9% and 28.5% of hospitals, respectively. This scarcity of resources is reflected in several other ‘field reports’ 6, 7.

There are currently no published data from sub-Saharan Africa on the ‘demand’ side of the equation; the proportion of patients who would be appropriate for critical care review and admission. A recent review of global critical care provision suggested important areas for future research, which included regional critical care needs assessment and availability of critical care resources 8. The aim of this study was to assess demand for critical care services at a large tertiary referral centre in Lusaka, Zambia. Additionally, we carried out a survey of the major medical institutions in Zambia that accepted acute admissions, in order to assess the absolute provision of critical care beds across the country.

The University Teaching Hospital (UTH) is the only tertiary referral centre for Zambia, and has an official inpatient capacity of 1655 beds 9. Services offered include adult and paediatric medicine and surgery, obstetrics and neonatal care. It has a 10-bed critical care unit, of which eight beds could be said to offer care comparable, although not matching, with level-3 care as defined by the UK Intensive Care Society 10, including mechanical ventilation and vasopressor support. Although haemofiltration is only occasionally available on the unit, haemodialysis is offered by a renal unit in the hospital.

This study investigated three acute units in UTH, which included the ‘filter’ ward (analogous to an emergency department major cases area), medical admission ward (MAW) and surgical admission ward (SAW). Anecdotally, the demand for critical care services far outstrips provision at UTH. The aim of the study was to quantify demand, using an objective scoring system for inpatient morbidity. Additionally, we wished to look at the diagnoses of patients requiring critical care, and particularly the prevalence of the Human Immunodeficiency Virus (HIV) rate among admitted patients, as the HIV rate in urban areas in Zambia is as high as 20% 11, and whether HIV positive patients were more likely to require critical care services.

Methodology

The study was approved by the University of Zambia Biomedical Research and Ethics Committee. Each patient gave informed verbal consent to be included in the study. Written consent was not required as no intervention was performed, the monitoring procedures were not harmful and were part of normal day-to-day care. Appropriate steps were undertaken to ensure patient confidentiality, including de-identification of data. We assessed all adult inpatient admissions to filter, MAW and SAW over a 48-h period covering 1–2 December 2014. Patients did not have their data recorded more than once. The following patients were not studied: obstetric patients; patients under 16 years; and patients not yet assessed by a clinician (doctor or non-physician medical clinical officer).

We conducted a prospective point prevalence study of morbidity in acute admissions to quantify demand for critical care services. This was compared with the UK Royal College of Physicians (RCoP) guidelines for critical care review and admission. The study comprised of a snapshot of the acute illness burden at the UTH using the National Early Warning Score (NEWS) 12. The NEWS is a scoring system developed by the RCoP as a national ‘track and trigger’ tool for assessing the deteriorating patient, and is endorsed by the National institute for Health and Care Excellence (NICE), UK 13. High scores are associated with higher mortality and a requirement for intensive care admission in the UK context. The NEWS measures six physiological parameters. Each parameter is awarded a score depending on the degree of divergence from physiologically normal values, which trigger a response when a defined threshold is reached (total score ranges from 0–19). These triggers should determine the urgency of the clinical response and who responds. For example, a low score (1–4) prompts a nurse-led review, while a medium score (5–6) prompts a physician review. A high score of seven or more should prompt ‘emergency assessment by a critical care outreach team’ who will ‘usually transfer the patient to a higher dependency care area’ 12 (Table 1).

Three members of the investigation team (all physician anaesthetists – authors PD, SM, ZR) independently assessed each six physiological parameters of the NEWS (respiratory rate, oxygen saturations, temperature, systolic blood pressure, pulse rate and level of consciousness (AVPU scoring system 14)) as well as the requirement for supplemental oxygen. As oxygen was not always available, we defined this as either currently receiving oxygen, or oxygen saturations ≤ 95% on room air. We further reviewed admission notes for the working diagnosis as detailed by the most senior clinician to have reviewed the patient. HIV status (where available), age and sex were also recorded. To ensure the study was not detrimental to patients care, we transcribed our own observations onto the nursing observation chart, where they were not recorded, and informed the team in charge of the patients’ care if the physical parameters suggesting further review was indicated. We collected data from 08:00 to 18:00 over the two data collection days. Patients were assessed only once during this period to avoid repetition, however, the time at which they were assessed was not fixed. Patients admitted overnight were identified from admission records and assessed the following day. No overnight admissions were missed. We used the following equipment for data collection: a portable oxygen saturations monitor (LifeBox, London, UK); portable sphygmomanometer (Bosch and Sohn, Stuttgart, Germany); and tympanic temperature probe (Braun, Meslungen, Germany).

To put the demand for services into context, we retrospectively reviewed admission rates for the month of November 2014. Data were taken from the ‘filter assessment log’, which held details of all admissions. No patient identifiable data were collected.

For information on service provision at UTH, we interviewed ward and ICU nursing leaders, as well as intensive care and medical consultants. To assess the current provision of critical care services across Zambia, we investigated 49 medical institutions with acute inpatient beds. The Zambian government had not previously recorded critical care beds as distinct from hospital beds in its routine health statistics. However, we were kindly assisted with data collection by the Zambian Ministry of Health, and 30 institutions were identified as tier two or three hospitals (secondary and tertiary referral hospitals, and larger primary institutions). Information was gathered by telephoning, emailing or physically visiting the institutions, where necessary. The appropriate medical teams were asked to list the total number of dedicated ICU and HDU beds at the facility, and the number of beds that could offer mechanical ventilation. The response rate was 100%.

Owing to the limited resources in Zambia, the Intensive Care Society definitions for level of care for a critical care bed were felt to be inappropriate. For example, a nursing ratio of one nurse to two patients was not consistently available in any public hospital nationally. As such, we defined a critical care bed as the self-reported number of dedicated intensive care or high dependency care beds, where higher nurse to patient ratios and more frequent observation could be delivered. A critical care bed offering mechanical ventilation was defined as one where a patient could receive mechanical ventilation without requiring transfer to another institution.

Data analysis included Fisher's exact test for categorical data and the Kruskal–Wallace test for ordinal data or continuous data that was skewed. A p < 0.05 was used as the level for significance. IBM SPSS Statistics software was used (Version 22; IBM, Armonk, NY, USA).

Results

We assessed 120 acute admissions over 48 h, which included 23 patients from the filter ward, 75 patients from MAW and 22 patients from SAW. As obstetric patients and patients under 16 years were seen in separate, dedicated obstetric and paediatric admission areas, respectively, it was not necessary to exclude any patients on this basis. However, at the end of the second day of data collection, seven patients were in the process of being assessed, and were not studied. The median age was 38.0 (28.0–48.5 [16,0–84.0]) years. Seventy-seven (64%) patients were male and 56 (36%) female. HIV test results were available for 75 patients (63%). Of those patients where a test had been performed, 53 (71%) were HIV positive (44% of all patients, including those without test results). The Joint United Nations Programme on HIV/ Acquired Immune Deficiency Syndrome (UNAIDS) has estimated Zambia's HIV prevalence to be 13.78% 11.

Fifty-four patients (45%) had a NEWS score of seven or more, which in the UK context would indicate a requirement for critical care review (and usually admission), according to RCoP guidelines. To estimate the annual total number of acute adult medical and surgical admissions to UTH, we collected complete admission data for November 2014. There were 1438 patients admitted, with a daily average of 48 patients, equating to an estimated 17,500 admissions in a 365-day year, assuming no monthly variation.

The patients’ median systolic blood pressure was 112 (100–132 [70–299]) mmHg, with 58 (48%) having a systolic blood pressure outside the normal physiological limits (as defined by NEWS). The median heart rate was 98 (80–111 [51–152]) beats.min⁻¹, with 74 (62%) patients having a heart rate outside normal physiological limits. Only 14 (12%) patients received oxygen, despite 62 (52%) patients potentially being hypoxic (oxygen saturations ≤ 95% on air, or receiving concurrent oxygen therapy), but this may be explained by the lack of availability of oxygen. Eighty-five (71%) patients had a respiratory rate outside of normal physiological range. On neurological assessment, 29 (24%) patients responded only to verbal or painful stimuli, or were unresponsive (Table 2).

NEWS were normally distributed for all patients, but showed a positive skew for the surgical cohort and a negative skew for filter patients. A NEWS score of 1–4 (defined as low) was observed in 37 (31%) patients, while a NEWS score of 5–6 (medium) was observed in 21 (18%). A NEWS score of seven or more (high) was recorded in 54 (45%) patients, which by RCoP guidelines requires prompt emergency assessment by a critical care clinical team, and usually transfer to a higher dependency care area (Table 4). The mean NEWS score for all patients was 6.34 (3.94), which varied by ward assessed; in the filter ward (n = 23), NEWS 9 (6–11 [0–15]); MAW (n = 75), NEWS 6 (3–9) [0–15]); and SAW (n = 22), NEWS 4.5 (0.25–5.75 [0–15]) (Table 3).

Table 4 illustrates the recorded diagnoses for medical and surgical admissions. A maximum of three diagnoses were recorded per patient, which were the working diagnoses by the most senior medical practitioner to have reviewed the patient. These were preliminary diagnoses usually based on clinical findings alone, before relevant investigations. The most common diagnosis was infection/sepsis, which accounted for 58 (48%) of all admissions. Other common diagnoses included sequelae of uncontrolled chronic co-morbidities; for example, pulmonary oedema/congestive cardiac failure (CCF) accounted for 11 (9%) medical admissions, and hypertensive crises accounted for 11 (9%) admissions. Only one (1%) medical admission had ischaemic heart disease listed as a diagnosis.

Although not part of the study proposal, due to the very high rate of HIV positivity observed, post-hoc analysis was performed on the incidence of diagnosed infectious disease in patients according to their HIV status, and of NEWS scores according to HIV status. HIV-positive status was associated with higher rates of presentation with infection (35/53, (66%)) compared with HIV-negative patients (4/22, (18%), p = 0.0003), or patients who had not been tested (19/45, (42%), p = 0.0003).

Put reciprocally, of the 39 patients with infectious pathology who had been tested for HIV, 35 (90%) were HIV positive. Although rates of HIV positivity were observed to be higher among hospital admissions than the national average (53/75 patients tested (71%) vs. 13.78% nationally 11), no statistically significant association was noted between HIV status and NEWS ≥ 7. There was no significant difference when comparing all NEWS scores for the categories of HIV status: HIV positive 6 (4–9 [0–14]), HIV negative 6.50 (2.25–10.75 [0–15]) and unknown 5 (3–9 [0–15]), p = 0.51.

The medical admissions ward had 21 male and 21 female beds, staffed by six nurses and an on-call medical team of six doctors. Despite having only 42 beds, there were 57 patients in MAW on the day of investigation, which was reportedly typical when discussed with the ward staff. The surgical admissions ward had 18 beds, staffed by three nurses and three junior doctors. As with the MAW, there were typically more patients than beds, with the extra patients occupying mattresses or blankets on the floor.

The ICU had 10 beds and could offer mechanical ventilation for a maximum of eight patients. The nurse-to-patient ratios were one nurse to two patients, at best, at the time of data collection. The ICU medical team consisted of two or three junior doctors (typically one or two trainee anaesthetists and a trainee physician), and a consultant during weekdays. At night, the ICU was covered by the on-call anaesthetic trainee. In terms of infrastructure and equipment, supplemental oxygen was supplied from cylinders, with the exception of the ventilators in intensive care, which were generally supplied by pipeline oxygen. The pipeline oxygen for the ICU was delivered from an on-site oxygen concentrator. However, the oxygen concentration delivered was unclear, and not routinely measured. There was the facility to administer supplemental oxygen to six beds in the filter unit, six beds in the MAW and to all ten beds in the ICU.

There were 91 reported adult intensive care or high dependency beds nationally, of which 66 could offer mechanical ventilation. There were also 18 paediatric and neonatal critical care beds nationally, of which 5 offered mechanical ventilation. Of the total of 109 critical care beds, 81 (74%) were government-funded, and of 74 beds offering mechanical ventilation, 58 (78%) were government-funded. There was a 100% response rate from the 49 institutions contacted (Fig. 1).

Discussion

This point prevalence study aimed to evaluate the inpatient morbidity of acute admissions to a tertiary referral hospital in Zambia using an objective tool, and to compare this with critical care provision. The NEWS scores indicated a marked inpatient morbidity burden, and a likely gross shortfall in provision of critical care beds. However, in the context of a severely resource-limited setting, this can be seen as only part of the problem.

Of 120 inpatients assessed, 54 (45%) had physiological derangement that would prompt critical care review ‘and usually admission’ in the UK context. Strikingly, HIV rates were extremely high among the patients we studied. Of inpatients who had been tested for HIV 53 (71%) were positive. Furthermore, 58 (48%) admissions were associated with infection/sepsis, and a large proportion of these patients were HIV positive (90% of those tested). This reflects the huge HIV burden experienced by Zambia. Latest UNAIDS data suggest a prevalence of 13.78%. In 2014 (latest available data), 2.4 million people in Zambia were tested for HIV 11. At UTH all admissions were tested for HIV within 48 h of admission. Antiretroviral therapy (ART) was increasingly available, and in 2012, 93% of adult patients living with HIV had had ART started. At the last survey, 80% of patients living with HIV initiated on treatment were still on treatment 12 months later. However, loss to follow-up at 12 months after initiation was as high as 47% in certain parts of Zambia.

On the supply side at UTH, there were eight critical care beds with the facilities for mechanical ventilaion. The UTH had an admission rate of approximately 17,500 patients per year. Extrapolated from the observed proportion of 45% of patients in the study, an estimated 7900 patients might be expected to have a NEWS of seven or more. Over the two-day period studied, this indicated a significant mismatch between supply and demand of critical care. This did not include obstetric, paediatric, postoperative and other inpatients, and did not account for the fact that most intensive care patients stayed for more than one day. The actual mismatch on a rolling basis is likely to be higher.

Nationally, we identified a total of 109 critical care beds including public, private health care, adult paediatric and neonatal units, of which 71 had the capability of supporting mechanical ventilation. In comparison, England has 3996 adult critical care beds 15 for a population of 53 million (7.5 beds/100,000 population) 16, while Zambia has 91 for 15 million people (0.61 beds/100,000 population) 17. Notwithstanding the limitations of this study, the shortfall in provision of critical care services is likely to be very large.

Zambia is classed as a lower middle-income country by the World Health Organisation (GDP per capita $1045–$12,746), but GDP has roughly trebled in a decade. Despite this rapid growth, health care infrastructure appears to be lagging behind the economic boom experienced in Zambia. Although increased provision of critical care beds is important, it may be difficult to achieve in the short to medium term, and is not a solution to a larger health care problem. The provision of more basic facilities such as oxygen in the acute wards, for example, may have a greater impact on acute illness, and may also reduce demand for critical care services.

There is no universally agreed model for what constitutes a critical care bed. However, the view that critical care is a ‘luxury’ specialty in resource-poor settings is increasingly contested. Indeed, the Lancet Commission on Global Surgery includes development of critical care as an essential aspect of improving overall peri-operative facilities 2. A previous study identified lack of availability of ICU beds and equipment as a possible contributing factor in avoidable peri-operative mortality at UTH 18. At the time of the study, the UTH ICU casemix showed a predominance of medical over surgical patients. Given the potentially avoidable nature of peri-operative morbidity and mortality compared with that of heart failure and other irreversible comorbidities, it may be argued that a surgical focus might lead to better outcomes from the resources available.

Our study showed that a high number of acutely-ill patients were HIV positive, presented with infection, and had a low median age. Although HIV infection appeared to be an independent risk factor for inpatient admission (71% of tested inpatients vs. 13.78% nationally), it was not associated with an increased chance of having a high NEWS score (≥ 7). Many of these patients with acute, reversible disease could reasonably be expected to benefit from access to timely treatment, effective resuscitation, appropriate antibiotics and antiretroviral therapy.

Careful patient selection is therefore clearly needed, and potentially contentious. The instigation of measures such as a locally appropriate NEWS system, with basic monitoring such as portable oxygen saturation monitors, may be important steps towards early identification of deteriorating patients. Similarly, the role of a critical care outreach team to assess the response to treatment and escalate patients appropriately has been an effective response to resource limitation in the UK, and may be of benefit in the Zambian context.

Our study has established the feasibility of using an objective, rapid and cheap measure of illness such as the NEWS system to provide a consistent approach to the triaging and care of hospital inpatients. Further work is needed to determine if a locally-appropriate system could be adapted to perform better in a resource-challenged setting such as Zambia.

There are several limitations to this study. The data on demand for services were collected from a single hospital, in contrast to the nationwide supply data. The UTH is Zambia's only tertiary centre, and serves a predominantly urban population. This may impact on the degree of delay before presentation in seeking health care, the severity of pathology at presentation, adequacy of pre-tertiary hospital care and other social and cultural factors related to seeking or accessing health care. The metrics for morbidity were limited to those required for NEWS, and more robust measurements of physiological derangement (such as mean arterial pressure and Glasgow coma score) may be better indicators of inpatient morbidity. In addition, the measurement of NEWS was taken at any time from patients’ arrival in the filter unit until 48 h post admission, and the score may have been changed by any treatments received during that period.

The data on supply of critical care beds nationally relied on self-reporting of the designated higher care beds. This is a subjective definition, and it is likely that there is heterogeneity in the descriptions of the level of care provided by different responders. Finally, we accept that while the NEWS score provides a useful measure of morbidity, it has not been validated in the Zambian context. In the UK setting, high NEWS scores have been shown to be predictive of ICU admission, cardiac arrest and death. This may not be the case at UTH. The patient population and burden of disease are different, and future research is needed to confirm that NEWS has predictive power in a low-resource context.

In conclusion, we have demonstrated a high burden and acuity of illness at a large tertiary hospital in Lusaka, Zambia. This coincides with a significant deficiency in the ‘supply’ of critical care beds, with only 109 critical care beds available nationally (including adult, paediatric, neonatal, private and public). Although Zambia is a rapidly-emerging southern African economy, state-funded and private health care still provide very limited access to critical care services for the population of 15 million people. The HIV rates in patients presenting with acute illness were also very high, often with other concomitant infection. As the economy and population of Zambia grows, careful planning of both hospital infrastructure and systems will be essential to increase access to critical care services. The conclusion that more critical care beds are needed is important for future service planning, but it is as, or perhaps more important, that basic health care resources are rapidly developed to potentially reduce the requirement for critical care.

Competing interests

No other external funding or competing interests declared.