ABSTRACT.
In much of sub-Saharan Africa, lumbar punctures (LPs) are performed less frequently than indicated. This is often attributed to patient/family refusal; however, other factors have not been systematically evaluated. We investigated predictors of LP performance for a prospective cohort of people with HIV and new-onset seizures at three hospitals in Zambia. We enrolled 257 participants, including 184 (72%) adults and 144 (56%) urban participants. LPs were performed for 65% of adults and 33% of children, and for 69% of urban and 38% of rural participants. In multivariate logistic regression analyses, LP completion was significantly less likely at one rural site and among children compared to adults. The worst WHO HIV disease stage was associated with increased odds of undergoing LP. Low LP completion rates in Zambia are multifactorial and related to health system and provider factors and patient/family preferences. Further research is necessary to understand this complex problem and develop interventions to improve LP uptake.
Lumbar puncture (LP) is a minimally invasive, generally safe procedure used to acquire cerebrospinal fluid (CSF) for laboratory analysis, and it is the standard of care for evaluating possible central nervous system (CNS) infections.1,2 Because clinical presentations of many CNS pathogens substantially overlap, a definitive diagnosis and treatment are not possible in the absence of CSF acquisition. When CSF is unavailable, clinicians initiate empiric antimicrobial therapy based on presumptive diagnoses. However, empiric treatment with broad-spectrum antibiotics, antivirals, antiretrovirals, antifungals, and antituberculosis treatment is prohibitive in terms of drug toxicities, interactions, and cost, especially among people with HIV, who are susceptible to opportunistic infections and usual CNS pathogens and in whom multiple infections may co-occur.3
Unlike other neurodiagnostic technologies such as neuroimaging and electroencephalography,4 basic CSF analyses are generally available at district-level or higher-level hospitals in lower-income and middle-income countries (LMICs).5 Despite this, substantial data indicate that LPs are often not completed when clinically indicated, especially in sub-Saharan Africa, where 50% of senior academic physicians consider patient/proxy refusal of an LP a major barrier to clinical care.6,7 The underlying determinants of poor uptake are unclear because the focus of most published studies is limited to queries of patient and proxy willingness to consent, thereby addressing only patient/proxy barriers. Given the high prevalence of CNS infections in sub-Saharan Africa, especially among people with HIV,8 improving LP uptake might offer substantial public health benefits. We hypothesized that LP completion is also associated with provider-level and health system-level factors; therefore, we evaluated this hypothesis within an ongoing prospective cohort study of people with HIV and new-onset seizure.
Within a prospective cohort study of people with HIV (both adult and children) and new-onset seizures at three hospitals in Zambia (two rural and one urban; two public facilities and one private facility), sociodemographic and clinical data were collected within 2 weeks of the index seizure. Research resources provided LP supplies to all sites and facilitated extensive, clinically relevant diagnostic studies if CSF had been obtained, but LP completion was at the discretion of the treating clinicians. Participants, their proxies, and clinicians were clearly informed about the additional testing available. Written informed consent was obtained from all participants or their proxies for enrollment in the cohort study. This study was approved by the University of Zambia Biomedical Ethics Review Council, the Zambia National Health Research Authority, and the University of Rochester Research Subjects Review Board.
For this study, we retrospectively analyzed associations between LP completion and clinical, demographic, and health system factors for the overall cohort and then stratified them by participant age (adult/pediatric) and setting (urban/rural). Wilcoxon rank-sum tests, t tests, and χ2 tests were used, as appropriate, to compare groups using univariate analyses. Variables with P < 0.20 were included in multivariable logistic regression models to identify independent predictors of LP completion. Variables that were not collected at all sites were excluded from multivariate analyses. A statistical analysis was completed using Stata 14.0 (StataCorp., College Station, TX). P < 0.05 was considered statistically significant.
Overall, 257 participants were enrolled; of these participants, 184 (72%) were adults and 144 (56%) were evaluated at an urban hospital (Table 1). Adults had a mean age of 38 ± 10 years, and 52% were male. Children had a mean age of 6.6 ± 4.6 years, and 53% were male. At enrollment, half of all participants were using antiretroviral therapy. The median CD4 count was 201 cells/uL (interquartile range [IQR], 68–439) and was highest among children (median, 410; IQR, 130–835). Seizures were focal (by history) for one-third of the participants. In-hospital mortality was 26% (N = 67) overall.
Table 1.
Demographic characteristics of the overall cohort and sub-cohorts
| Demographic characteristics | Overall cohort (N = 257) | Adults (N = 184) | Children (N = 73) | Urban (N = 144) | Rural (N = 113) |
|---|---|---|---|---|---|
| Site, n (%) | |||||
| Rural 1 | 19 (7%) | 15 (8%) | 4 (5%) | 19 (17%) | |
| Rural 2 | 94 (36%) | 74 (40%) | 20 (27%) | 94 (83%) | |
| Urban | 144 (56%) | 95 (52%) | 49 (67%) | 144 (100%) | |
| Urban, n (%) | 144 (56%) | 95 (52%) | 49 (67%) | ||
| Pediatric participants, n (%) | 73 (28%) | 49 (34%) | 24 (21%) | ||
| Male, n (%) | 135 (52%) | 96 (52%) | 39 (53%) | 81 (56%) | 54 (48%) |
| Age, mean (SD) | |||||
| Adults | 37.8 (10.3) | 37.8 (10.3) | 37.0 (10.2) | 38.8 (10.4) | |
| Pediatrics | 6.6 (4.6) | 6.6 (4.6) | 6.9 (4.9) | 6.2 (4.1) | |
| LP performed during admission, n (%) | 143 (56%) | 199 (65%) | 24 (33%) | 100 (69%) | 43 (38%) |
| HIV characteristics | |||||
| Using ART, n (%) | 133 (52%) | 97 (53%) | 36 (49%) | 56 (39%) | 77 (68%) |
| CD4 count, cells/uL, median (IQR) | 201 (68–439) | 162 (64–364) | 410 (130–835) | 162 (54–393) | 294 (115–534) |
| Log HIV viral load, mean (SD) | 9.2 (3.7) | 8.3 (3.8) | 10.3 (3.2) | 10.1 (3.1) | 8.9 (3.8) |
| Virally suppressed at enrollment, n (%) | 32 (23%) | 25 (29%) | 7 (13%) | 4 (13%) | 28 (26%) |
| WHO stage at enrollment, n (%) | |||||
| Stage I | 61 (24%) | 47 (26%) | 14 (19%) | 19 (13%) | 42 (37%) |
| Stage II | 11 (4%) | 7 (4%) | 4 (5%) | 5 (4%) | 6 (5%) |
| Stage III | 70 (27%) | 59 (32%) | 11 (15%) | 37 (26%) | 33 (29%) |
| Stage IV | 113 (44%) | 69 (38%) | 44 (60%) | 81 (57%) | 32 (28%) |
| Time from HIV diagnosis to ART initiation, days, median (IQR) | 80 (7–182) | 49 (5–182) | 100 (22–182) | 81 (0–182) | 78 (8–182) |
| Time from ART initiation to index seizure, days median (IQR) | 582 (10–1,752) | 897 (10–2,321) | 266 (12–1,500) | 266 (47–2,515) | 817 (2–1,707) |
| Time from HIV diagnosis to seizure, days, median (IQR) | 155 (2–1,623) | 177 (6–1,774) | 100 (0–1,542) | 82 (0–471) | 1008 (12–2,111) |
| Time from HIV diagnosis to study enrollment, days median (IQR) | 160 (7–1,624) | 178 (12–1774) | 102 (3–1548) | 86 (6–473) | 1011 (19–2,112) |
| Clinical characteristics | |||||
| Time between seizure and enrollment, days, median (IQR) | 2 (1–4) | 1 (1–3) | 3 (1–5) | 3 (2–6) | 1 (0–3) |
| No. of seizures during 2 weeks prior to enrollment, median (IQR) | 2 (1–4) | 2 (1–4) | 2 (1–3) | 2 (1–2) | 2 (1–4) |
| History of focal seizures, yes, n (%) | 86 (33%) | 54 (29%) | 32 (44%) | 70 (49%) | 16 (14%) |
| Symptoms preceding seizure, n (%) | |||||
| Headache | 97 (60%) | 63 (71%) | 34 (46%) | 23 (47%) | 74 (65%) |
| Fever | 109 (42%) | 58 (32%) | 51 (70%) | 78 (54%) | 31 (27%) |
| Focal weakness | 24 (15%) | 13 (15%) | 11 (15%) | 8 (16%) | 16 (14%) |
| Meningismus | 30 (18%) | 17 (19%) | 13 (18%) | 11 (22%) | 19 (17%) |
| Encephalopathy | 61 (38%) | 46 (52%) | 15 (20%) | 12 (24%) | 49 (43%) |
| Medical history, n (%) | |||||
| Cerebral malaria | 4 (2%) | 3 (3%) | 1 (1%) | 1 (2%) | 3 (3%) |
| Coma | 10 (6%) | 9 (10%) | 1 (1%) | 1 (2%) | 9 (8%) |
| Loss of consciousness | 8 (5%) | 7 (8%) | 1 (1%) | 0 (0%) | 8 (7%) |
| Meningitis | 5 (3%) | 3 (3%) | 2 (3%) | 2 (4%) | 3 (3%) |
| Stroke | 7 (4%) | 5 (6%) | 2 (3%) | 2 (4%) | 5 (4%) |
| Opportunistic infection | 49 (30%) | 31 (35%) | 18 (25%) | 14 (29%) | 35 (31%) |
| Hospitalization characteristics | |||||
| Temperature on admission, mean (SD) | 36.9 (1.5) | 36.6 (1.3) | 37.3 (1.5) | 36.8 (1.5) | |
| Fever on admission, n (%) | 34 (21%) | 12 (13%) | 22 (31%) | 25 (22%) | |
| Glasgow Coma Scale (total score) on admission, median (IQR) | 15 (11–15) | 15 (14–15) | 12 (6–15) | 15 (13–15) | 15 (11–15) |
| Length of hospitalization, days, median (IQR) | 11 (5–17) | 8 (3–17) | 13 (8–17) | 14 (10–20) | 8 (3–16) |
| Died before discharge, n (%) | 67 (26%) | 44 (24%) | 23 (32%) | 40 (28%) | 27 (24%) |
ART = antiretroviral therapy; IQR = interquartile range; LP = lumbar puncture.
LP was completed for 119 (65%) adults, 24 (33%) children, 100 (69%) urban participants, and 43 (38%) rural participants (Table 1). In the univariate analyses, one rural site was less likely to complete LPs than the other two sites in the overall, adult, and rural cohorts, although there was no significant difference by site for the pediatric cohort (Table 2, Supplemental Table 1). Pediatric participants were less likely to have an LP completed than adults in the overall and rural cohorts. Other differences in HIV, clinical, and hospitalization characteristics were noted and are detailed in Table 2 and Supplemental Table 1.
Table 2.
Comparison of demographic, HIV, clinical, and hospitalization characteristics of participants who did and did not undergo complete lumbar puncture in the overall cohort and among adult, pediatric, urban, and rural sub-cohorts
| Demographics | Overall cohort | ||
|---|---|---|---|
| LP completed (N = 143) | LP not completed (N = 114) | P | |
| Site, n (%) | |||
| Rural 1 | 12 (8%) | 7 (6%) | < 0.001 |
| Rural 2 | 31 (22%) | 63 (55%) | |
| Urban | 100 (70%) | 44 (39%) | |
| Urban, n (%) | 100 (70%) | 44 (39%) | < 0.001 |
| Male, n (%) | 75 (52%) | 60 (53%) | 1 |
| Pediatric participants, n (%) | 24 (17%) | 49 (43%) | < 0.001 |
| Age, mean (SD | |||
| HIV characteristics | |||
| Using ART, [n (%) | 68 (48%) | 65 (57%) | 0.13 |
| CD4 count, cells/uL, median (IQR) | 178 (66–404) | 272 (100–534) | 0.06 |
| Log HIV viral load, mean (SD) | 8.9 (4.0) | 9.3 (3.5) | 0.54 |
| Virally suppressed at enrollment, n (%) | 14 (25%) | 18 (21%) | 0.58 |
| WHO stage at enrollment, n (%) | |||
| Stage I | 27 (19%) | 34 (30%) | 0.047 |
| Stage II | 6 (4%) | 5 (4%) | |
| Stage III | 35 (25%) | 35 (31%) | |
| Stage IV | 73 (52%) | 40 (35%) | |
| Time from HIV diagnosis to ART initiation, days, median (IQR) | 144 (22–182) | 77 (3–181) | 0.26 |
| Time from ART initiation to index seizure, days, median (IQR) | 1,168 (231–2,338) | 200 (4–1,580) | 0.03 |
| Time from HIV diagnosis to seizure, days, median (IQR) | 161 (4–1,636) | 143 (1–1,594) | 0.78 |
| Time from HIV diagnosis to study enrollment, days, median (IQR) | 164 (10–1,636) | 150 (6–1,595) | 0.56 |
| HIV diagnosis within 30 days of study enrollment, n (%) | 46 (37%) | 43 (38%) | 0.8 |
| Medical history, n (%) | |||
| Cerebral malaria | 2 (3%) | 2 (2%) | 0.62 |
| Coma | 9 (15%) | 1 (1%) | 0.001 |
| Loss of consciousness | 4 (6%) | 4 (4%) | 0.48 |
| Meningitis | 5 (8%) | 0 (0%) | 0.007 |
| Stroke | 4 (6%) | 3 (3%) | 0.43 |
| Opportunistic infection | 15 (24%) | 34 (34%) | 0.19 |
| Seizure characteristics | |||
| History of seizures, n (%) | 5 (4%) | 5 (4%) | 0.75 |
| Time between seizure and enrollment, days, median (IQR) | 2 (1–3) | 2 (1–4) | 0.81 |
| Number of seizures in 2 weeks prior to enrollment, median (IQR) | 2 (1– 3) | 2 (1–4) | 0.36 |
| History of focal seizures, n (%) | 50 (35%) | 36 (32%) | 0.57 |
| Symptoms preceding seizure [n (%)] | |||
| Headache | 43 (69%) | 54 (54%) | 0.053 |
| Fever | 60 (42%) | 49 (43%) | 0.87 |
| Focal weakness | 14 (23%) | 10 (10%) | 0.03 |
| Meningismus | 18 (29%) | 12 (12%) | 0.007 |
| Encephalopathy | 30 (48%) | 31 (31%) | 0.03 |
| Hospitalization characteristics | |||
| Temperature on admission, mean (SD) | 36.6 (1.4) | 37.0 (1.5) | 0.13 |
| Fever on admission, n (%) | 7 (11%) | 27 (27%) | 0.02 |
| Glasgow Coma Scale (total score) on admission, median (IQR) | 15 (14–15) | 15 (9–15) | 0.001 |
| Length of hospitalization, days, median (IQR) | 12 (6–18) | 8 (3–16) | 0.09 |
| Died before discharge, n (%) | 35 (24%) | 32 (28%) | 0.51 |
ART = antiretroviral therapy; LP = lumbar puncture.
Data from the urban adult cohort were collected before data were collected from the remaining cohorts. Therefore, some clinical data are not available for the urban cohort.
The multivariable logistic regression analyses of the overall cohort showed that LP completion was significantly less likely at one rural site (odds ratio [OR], 0.23; 95% CI, 0.07–0.70; P = 0.01) compared with other sites and among pediatric participants compared to adult participants (OR, 0.19; 95% CI, 0.09–0.39; P < 0.001) (Table 3). Each one-stage increase in WHO HIV disease stage at enrollment was associated with an increased odds of LP completion (OR, 1.31; 95% CI, 1.01–1.70; P = 0.04). In the urban cohort, pediatric participants (OR, 0.06; 95% CI, 0.02–0.23; P < 0.001) and those with higher WHO HIV disease stages at enrollment (OR, 2.02; 95% CI, 1.24–3.32; P = 0.005) were associated with LP completion. Other predictors were identified in the urban, rural, pediatric, and adult sub-cohorts and are presented in Supplemental Table 2.
Table 3.
Multivariable model of factors associated with LP completion in the overall cohort
| Overall cohort | |||
|---|---|---|---|
| OR | 95% CI | P | |
| Rural 1 (ref. site) | |||
| Rural 2 | 0.23 | 0.07–0.70 | 0.01 |
| Urban | 1.38 | 0.43–4.47 | 0.59 |
| Pediatric participants | 0.19 | 0.09–0.39 | < 0.001 |
| Using ART | 1.06 | 0.57–1.98 | 0.84 |
| WHO stage at enrollment* | 1.31 | 1.01–1.70 | 0.04 |
| Glasgow Coma Scale score at enrollment† | 1.07 | 0.98–1.17 | 0.1 |
ART = antiretroviral therapy; OR = odds ratio.
OR pertains to every one-stage increase in the WHO HIV disease stage.
OR pertains to every 1-point increase in the Glasgow Coma Scale score.
In our cohort of people with HIV and new-onset seizures in Zambia, LP completion was performed at the discretion of the clinical care team and, per usual care, required participant or proxy consent. As part of study participation, participants and treating clinicians were informed that additional clinically relevant CSF diagnostics not routinely available in the hospital laboratories were available free of charge to study participants. Despite this, the overall LP completion rate was low (56%).
Previous studies of LP uptake in sub-Saharan Africa, including Zambia9 and Botswana,10 as well as other regions11–15 have primarily focused on the patient’s or the parent’s/proxy’s willingness to consent to LP. However, our study investigated clinical and health system factors and identified several variables associated with LP completion. Most notably, one rural hospital was driving the differential LP rates observed for the overall, rural, and adult cohorts. The difference between this rural site and the urban site could be partially explained by differing ethnic backgrounds, education levels of patients and providers, and other cultural factors; however, the difference between the two rural sites could not be explained by these factors. The two participating rural sites are in the same Zambian province and serve the same ethnic population with similar educational attainment and socioeconomic levels. Among pediatric participants from the rural sites, there was no significant difference in the highest level of maternal education achieved or in overall household wealth between the two rural sties. Therefore, patient/family preference would not be expected to differ significantly between these two hospitals and cannot account for the difference in LP uptake observed between these two sites.
However, these hospitals do differ in their clinician staff and clinical volume. The rural site with greater LP uptake is a 200-bed hospital with approximately 4,000 admissions per year that is staffed by five physicians, 14 clinical officers, and one medical licentiate. Physicians provide all inpatient care at this hospital, and nonphysician healthcare workers provide outpatient care. The rural site with lower LP uptake is a 274-bed hospital with approximately 12,650 admissions per year that is staffed by five physicians who have completed training, 20 intern physicians, 14 clinical officers, and five medical licentiates. In contrast to the other rural site, inpatient care is provided by both physicians and nonphysician healthcare workers. Also of note was the markedly lower rates of LP completion among pediatric participants in the overall and urban cohorts but not the rural cohorts. In the rural setting, the same group of clinicians cares for both adult and pediatric patients, whereas the urban site has separate adult and pediatric hospitals with entirely separate care providers. This further suggests that provider-level factors also contribute to LP nonperformance in settings with fewer resources, especially those with physician shortages. Further investigation is therefore warranted.
Several clinical factors, including WHO HIV disease stage at enrollment, preceding symptoms, and focal seizures, were also associated with LP completion. Participants with more advanced WHO HIV disease stages were more likely to undergo LP, suggesting that clinicians may be more suspicious of an infectious etiology in these patients. Alternatively, patients/proxies may be more willing to consent to LP when the patient appears more ill; however, our data could not differentiate between these possibilities. Especially among pediatric participants, preceding symptoms suggestive of meningitis were associated with LP completion. Notably, pediatric participants with focal seizures were less likely to undergo LP than those who were described to have generalized seizures based on their clinical history. Because of the complete (rural sites) and relative (urban site) unavailability of neuroimaging in this setting, concern regarding mass lesions may have reduced LP completion in this clinical scenario; therefore, greater access to neuroimaging may improve LP completion in these settings. However, LP completion was not significantly different (P = 0.85) for the subset of participants who underwent neuroimaging and had no contraindications to LP compared with those who did not undergo neuroimaging.
This study had several limitations. Most notably, it had a retrospective nature that limited the analysis of important factors such as whether an LP was requested but consent was refused by the patient/proxy and whether clinician concerns regarding LP contraindications contributed to LP nonperformance for individual participants. Importantly, because this was an exploratory and hypothesis-generating study, we did not correct for multiple comparisons. Therefore, some of the significant associations observed, especially in the sub-cohorts with smaller sample sizes, likely occurred by chance. Still, results of this exploratory study indicate that participant, provider, and larger health system factors beyond an individual’s willingness to consent likely contribute to low LP completion in Zambia and, potentially, similar settings in sub-Saharan Africa. Further prospective evaluations are needed to confirm the findings of this study and examine a wider range of factors contributing to LP uptake to design locally contextualized interventions to improve LP uptake and, potentially, clinical outcomes for individuals with CNS infections in these settings.
Supplemental Material
Note: Supplemental tables appear at www.ajtmh.org.
REFERENCES
- 1. Greenlee JE, 1990. Approach to diagnosis of meningitis. Cerebrospinal fluid evaluation. Infect Dis Clin North Am 4: 583–598. [PubMed] [Google Scholar]
- 2. Scheld WM, Whitley RJ, Marra CM, 2014. Infections of the Central Nervous System. 4th ed. Philadelphia, PA: Wolters Kluwer Health. [Google Scholar]
- 3. Siddiqi OK. et al. , 2017. New-onset seizure in HIV-infected adult Zambians: a search for causes and consequences. Neurology 88: 477–482. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. World Health Organization, World Federation of Neurology , 2017. ATLAS: Country Resources for Neurological Disorders. 2nd ed. Geneva, Switzerland: WHO. [Google Scholar]
- 5. Parry E, Godfrey R, Mabey D, Gill G, 2004. Principles of Medicine in Africa. 3rd ed. Cambridge, UK: Cambridge University Press. [Google Scholar]
- 6. Herbert G, Ndiritu M, Idro R, Makani JB, Kitundu J, 2006. Analysis of the indications for routine lumbar puncture and results of cerebrospinal fluid examination in children admitted to the paediatric wards of two hospitals in East Africa. Tanzan Health Res Bull 8: 7–10. [DOI] [PubMed] [Google Scholar]
- 7. Thakur KT. et al. , 2015. Lumbar puncture refusal in sub-Saharan Africa: a call for further understanding and intervention. Neurology 84: 1988–1990. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Global Burden of Disease Meningitis Collaborators , 2018. Global, regional, and national burden of meningitis, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet Neurol 17: 1061–1082. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Hampande ML, 2009. Attitude and Determinant of Lumbar Puncture Acceptability Among Caregivers at Three Hospitals in Zambia. Lusaka, Zambia: University of Zambia School of Medicine. Available at: http://dspace.unza.zm/handle/123456789/551. [Google Scholar]
- 10. King MB, Rwegerera GM, 2015. An audit of consent practices and perceptions of lumbar puncture, Botswana inpatient setting experience. Afr J Emerg Med 5: 66–69. [Google Scholar]
- 11. Deng CT, Zulkifli HI, Azizi BH, 1994. Parents’ views of lumbar puncture in children with febrile seizures. Med J Malaysia 49: 263–268. [PubMed] [Google Scholar]
- 12. Khakshour A, Hashemi M, Tavakoli H, Sheikhi Z, Kiani MA, Elmi S, 2013. Evaluation of parental attitudes toward lumbar puncture in their children. Int J Pediatr 1: 19–23. [Google Scholar]
- 13. Alnajim SA. et al. , 2017. Common misconceptions of lumbar puncture complications among parents in the eastern region of the kingdom of Saudi Arabia. Int J Adv Res (Indore) 5: 1257–1264. [Google Scholar]
- 14. Narchi H, Ghatasheh G, Al Hassani N, Al Reyami L, Khan Q, 2012. Why do some parents refuse consent for lumbar puncture on their child? A qualitative study. Hosp Pediatr 2: 93–98. [DOI] [PubMed] [Google Scholar]
- 15. Narchi H, Ghatasheh G, Hassani NA, Reyami LA, Khan Q, 2013. Comparison of underlying factors behind parental refusal or consent for lumbar puncture. World J Pediatr 9: 336–341. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.