Eligibility Criteria

The populations of interest were civilians in LMICs affected by humanitarian crises, defined here as events stemming from armed conflicts, natural disasters, or food insecurity that threaten the health and safety of a community. These included populations remaining in areas affected by crises and those forcibly displaced from them as refugees and IDPs. Studies that focused on current or former military populations were excluded. High-income countries were excluded as the vast majority of humanitarian crises occur in LMICs and the resources available to tackle NCDs in LMICs are very different to those in high-income countries. The time periods of humanitarian crises included acute, chronic, and early recovery time periods.

The interventions of interest were health interventions covering health promotion, prevention, treatment, or rehabilitation activities at the individual or population level specifically for outcomes of NCDs.

The outcomes included morbidity/mortality due to NCDs and surrogate outcomes (e.g. blood pressure, blood glucose levels) at the individual or population level. In addition, we also included information on process outcomes (e.g. adherence to clinical treatment) and feasibility of interventions and measurement methods, if the study included data on changes in health outcomes. We did not include mental health outcomes as these have been reviewed elsewhere.[13]

Information Sources and Search Strategy

The following bibliographic databases were searched: MEDLINE, Embase, Global Health, PsychInfo, and IBSS. The search terms were: (i) disaster-related terms; AND (ii) research study-related terms; AND (iii) geographic terms; AND (iv) NCD terms. A search of the grey literature was also conducted across a range of humanitarian-related databases and standard search databases such as Google. The full search strategy is provided in S1 File. Studies published in any language between January 1980 and June 2014 were included.

Study Selection and Data Extraction

Citations from the search results were imported from the bibliographic databases into EndNote for screening for eligibility based on the eligibility criteria given above. Duplicates were removed and the remaining citations assessed by title or abstract, and a full text review then conducted. References of the remaining studies selected after the full text review were examined for potentially relevant articles based on the eligibility criteria. Analysis of the final selected studies was then conducted. This involved extracting data from the final selected studies into an Excel database, with key extraction variables including: author and date of publication, geographic setting, sample population characteristics, study objectives, NCD condition studied, intervention characteristics, outcomes measured, results of the intervention, study conclusions, study design, and quality. The data screening and extraction were conducted independently by two authors and any variances resolved between them.

Quality assessment

A quality assessment was conducted, with the Newcastle-Ottawa Quality Assessment Scale (NOS) version for cohort studies used for the observational studies [14, 15]. This was selected as it is a convenient and widely used tool with proven validity and reliability and has been endorsed by Cochrane Reviews [15–17]. For the randomised controlled trial (RCT) study we applied the widely used and validated Cochrane Risk of Bias Tool[18].

The NOS assigns stars for methodological rigour based on three categories: study selection, comparability of study groups, and outcome assessment. Studies were initially assessed within each category using the coding manual for cohort studies provided by Wells et al (see [15] and http://www.ohri.ca/programs/clinical_epidemiology/nosgen.pdf), with letters and descriptions assigned describing how each study fulfilled each criterion. Stars were then assigned per the NOS assessment scale when the study achieved high quality within that category. Criteria which were not applicable to particular studies were listed as not applicable but factored into overall impressions regarding that study’s conclusions.

The Cochrane Risk of Bias Tool was developed to promote the assessment of quality of trials based on their risk of biased conclusions rather than focussing on reporting and methodological constraints [18]. The RCT was therefore evaluated as being at either high, low, or unclear risk of bias in several domains (selection, performance, detection, attrition, reporting, and other bias), and a descriptive justification of each conclusion was provided.

Neither NOS nor the Cochrane Risk of Bias Tool uses an established summary score or threshold of quality, with the quality assessment primarily used to assess strengths and weaknesses of each study rather than to rank studies or to screen them out.

Synthesis of results

As the studies were heterogeneous in setting, intervention, and outcome, single effectiveness summary statistics across studies were not considered appropriate and were not estimated. Instead, a descriptive analysis of study results was reported.

Study Selection

The bibliographic databases yielded 4919 citations after duplicates were removed. Of these, only 8 met the study inclusion criteria (Fig 1).[19–26] The main reasons for excluding the 4879 studies were they were: in high-income countries; not in humanitarian contexts or took place too long after a humanitarian crises; not intervention studies; did not report changes in health outcomes; or were not full papers (e.g. conference abstracts only). These reasons applied at each screening stage. Exploring references from these 8 studies did not reveal any further studies meeting eligibility criteria. No studies were identified in the grey literature.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Results of screening process.

https://doi.org/10.1371/journal.pone.0138303.g001

Study Characteristics

Key characteristics of the final 8 selected studies are included in Table 1. The studies were published between 1997 and 2014, with 6 out of 8 published within the past five years. Seven of the studies were with populations affected by armed conflict, and the remaining study with a population affected by an earthquake.[25] Sample sizes of the study populations ranged from 28 patients included in the RCT [26] to 12,550 patients in a diabetes cohort study.[24] The studies were conducted in 5 different countries: Afghanistan,[19] Georgia,[21] India,[26] Jordan,[20, 22–24] and Turkey.[25] Four of the studies[20, 22–24] came from the same research group studying Palestinian refugees in Jordan, with their studies covering diabetes and hypertension. These four studies and one other[26] took place in long-term, relatively stable refugee settings, while the remaining studies were in more acute- or early post-crisis settings.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of studies examining effectiveness of interventions targeting NCDs during humanitarian crises.

https://doi.org/10.1371/journal.pone.0138303.t001

Of the 8 studies, 7 used observational study designs[19–25] and 1 was an RCT.[26] The observational studies consisted of 5 cohort designs,[20–24, 26] 1 case series[19] and 1 interrupted time series.[25] The RCT[26] was the only study to compare outcomes between two groups. None of the studies examined the cost of implementing the intervention or the cost-effectiveness of the intervention.

The studies examined a broad range of NCD conditions: arthritis,[26] chronic kidney disease,[25] diabetes,[22–24] heart failure,[21] hypertension,[20] and thalassaemia.[19] All studies examined outcomes at the individual patient level and were primarily focused on disease management rather than prevention or health promotion. Details of each intervention and key outcome measures, study results, and specific study conclusions are presented in Table 1.

Quality Assessment

The quality assessment identified a number of common weaknesses. The observational studies (assessed using NOS) were generally adequate in describing the study population and establishing exposure. Deficiencies common to the observational studies were predominantly related to comparability and follow-up. None had a defined comparison group or unexposed cohort. Study transparency was also noted to be a weakness common to the observational studies. No study addressed potential biases, nor did any study discuss how missing data were handled. Only three of the observational studies adequately reported follow-up periods of participants, and most studies inadequately described their follow-up procedures. Follow-up periods ranged from undefined[19] to three years[23], with studies from settings of chronic crisis demonstrating longer follow-up. Outcome assessment was also problematic. Most studies only provided self-reported outcomes; the outcomes reported by the Khader et al. papers were of slightly higher quality in that they were linked to electronic medical records, but those assessments were not described in a standardised way such as via the International Classification of Disease (ICD) codes. Only four studies[20–22, 24] partially discussed study limitations, and only one[23] gave any information on sources of funding, and even then only in the online version of the article.

The RCT study from India[26] was assessed using the Cochrane Risk of Bias Tool and was judged to have a high risk of selection, performance, and detection bias, primarily due to the study’s lack of blinding. The study used an open enrolment process and all members of the research team appear to have had knowledge of patients’ treatment. Although outcome reporting was a strength of the study, there was an overall high risk of additional biases given that a single non-blinded researcher assessed the outcomes. Further deficiencies surrounded the reporting of the randomisation process, which was not described in any detail. Further details on the scoring for individual studies are given in S2 File and S3 File.

Synthesis of results

Cardiovascular diseases were assessed via two cohort studies–one in Georgia examining a heart failure disease management programme[21] and one in Jordan examining hypertension care among Palestinian refugees.[20] The contexts of these two studies differed in the sense that the study in Georgia was examining the effectiveness of a health programme that then experienced the outbreak of war during the intervention, while the study in Jordan took place in a long-term refugee setting that was relatively stable during the study period. Both studies focused on the implementation of disease management algorithms in settings of humanitarian crisis and attempted to highlight both the feasibility and challenges of such programmes.

In the study in Georgia by Hebert et al.,[21] the heart failure disease management programme saw some success among its 400 patients by demonstrating a statistically significant increase in ejection fraction—the fraction of blood volume exiting the heart’s ventricles with each heartbeat, which tends to decrease in the most common types of heart failure—and a statistically significant decrease in blood pressure over the course of the 2-year programme. Ejection fraction improved by 4.1±2.6% (p<0.001) and systolic and diastolic blood pressures decreased by 30.9±20.0 mmHg and 17.8±13.0 mmHg, respectively (p<0.001 for both). The intervention also demonstrated a decrease in smoking rates and in emergency room visits and hospitalisations. Heart failure classification also improved.

The study by Khader et al. on a hypertension management programme in Jordan had mixed clinical results, with approximately three quarters of patients meeting blood pressure targets.[20] The intervention focused on the method of cohort monitoring by using an electronic medical record system to enrol patients in a cohort that could be studied over time. The monitoring allowed researchers to also assess if goals of care were being met, both with respect to hypertension care goals such as blood pressure monitoring and with respect to associated diseases such as hypercholesterolaemia and diabetes. Results were mixed; among a sub-cohort of 226 patients assessed for 12–15 months, 100% were screened for high cholesterol but only 3% were screened for diabetes using a glucose blood test. The study authors concluded that the interventions were an improvement on baseline care in both settings. However, no comparison group was included.

Three cohort studies, all by Khader et al., focused on diabetes care among Palestinian refugees in long-term refugee settings in Jordan.[22–24] All three studies conducted very similar interventions consisting of a standardised diabetes protocol and assessment of patient outcomes and programme outputs via electronic medical records-based cohort monitoring. The concept was very similar as well to the aforementioned study targeting hypertension in a similar patient population with the initial DM study essentially mirroring that design.[20] The subsequent two diabetes studies differed in terms of follow-up and scope, with one study[23] focusing on 12-, 24-, and 36-month outcomes, and the other on the expansion of the programme from one clinic to six.[24] While the Khader et al. studies had similar designs and settings, it was confirmed via correspondence with the studies’ authors that the populations of each study differed. For this reason—and because this review featured descriptive analysis rather than meta-analysis—it was felt that inclusion of each study for analysis was appropriate.

In general, the diabetes studies claimed an improvement in the programme over time. Earlier assessment of the programme [22] found that over half of patients were not receiving post-prandial blood glucose checks and that those checked only demonstrated proper diabetes control (</ = 180mg/dl) half of the time. Subsequent assessment described in 2014[24] found that most programme outputs had improved, with nearly all patients attending clinic meeting the blood testing goals. However, other treatment goals, specifically foot examination and ophthalmology referral, that were problematic during the earlier study did not continue to be assessed in the subsequent studies. These studies also found that loss to follow-up rose over time (Γ² test-for-trend = 43.5; p<0.001). Nevertheless, the study authors contend that having a monitored cohort using an electronic medical record-based system could allow for improved retention of patients through more proactive patient monitoring.

Chronic kidney disease was assessed by one retrospective study from Sever et al. in the Marmara region of Turkey in the aftermath of an earthquake.[25] This study focused on both the infrastructure changes and clinical patient outcomes of providing haemodialysis to patients with severe chronic kidney disease. The study found that infrastructure for providing haemodialysis was affected by the earthquake, with an acute decrease in haemodialysis centres, machines available, personnel, and subsequently numbers of haemodialysis treatments provided. Gradually these numbers improved during follow-up. The initial earthquake also led to an increase in the number of patients receiving once-weekly (i.e., less frequent) haemodialysis.

Despite the infrastructure challenges, the authors found that mean interdialytic weight gain—the amount of weight patients gain between treatments, typically fluid weight due to poor blood filtration and urine production—actually decreased from a pre-earthquake 2.9±1.1 kg to 2.6±1.1 kg 1-week post-earthquake, despite the increased numbers of patients receiving haemodialysis less frequently. Moreover, the blood pressures of patients studied remained stable throughout the study period. The authors contend that adequate patient education regarding disaster preparedness and fluid restriction likely helped mitigate poor patient outcomes, although no comparison between the baseline health status of patients able to seek care after the earthquake and the status of the larger number of patients receiving haemodialysis before the earthquake was conducted.

The only RCT eligible for inclusion in this systematic review studied changes in limb mobility among 14 matched pairs of arthritis patients living in a stable Tibetan refugee setting in northern India.[26] In this open, non-blinded RCT, patients were randomised to receive three months of either traditional Tibetan arthritis treatments (herbal pills, dietary restriction, and behavioural advice) or Western medication (ibuprofen or indomethacin). In 12 of 14 pairs, the traditional Tibetan treatment led to greater improvement in limb mobility, and in the remaining 2 pairs the treatments performed equally well. Although they have not presented the data, the authors do suggest that pain control was better with the Western treatment than the traditional treatment. The authors state that a secondary objective of the study was to examine the process of conducting an RCT on traditional treatment options, although the authors do not comment specifically on the nuances of conducting an RCT in unstable settings.

One case series study by Bolt et al. examined thalassaemia among 45 paediatric Afghan patients seeking care at a United States-managed military hospital in a chronic crisis setting in Afghanistan.[19] The research team provided the intervention—palliative splenectomy—with the rationale that more curative treatment (e.g. stem-cell transplantation) would not be feasible in the Afghan context. The study reported an improvement in anaemia with mean haemoglobin levels rising from 5.4g/L pre-operatively to 8.7g/L post-operatively. Furthermore, frequency of blood transfusion decreased from every 24 days to approximately every 50 days before and after surgery. The authors state that families were pleased with the improvements during follow-up, although patient-specific data, confidence intervals, and specifics regarding follow-up were not provided.