Generalized seasonal autoregressive integrated moving average models for count data with application to malaria time series with low case numbers

Application of improved harmonic Poisson segmented regression model in evaluating the effectiveness of Kala-Azar intervention in Yangquan City, China

Background

The Centre for Disease Control and Prevention in Yangquan, China, has taken a series of preventive and control measures in response to the increasing trend of Kala-Azar. In response, we propose a new model to more scientifically evaluate the effectiveness of these interventions.

Methods

We obtained the incidence data of Kala-Azar from 2017 to 2021 from the Centre for Disease Control and Prevention (CDC) in Yangquan. We constructed Poisson segmented regression model, harmonic Poisson segmental regression model, and improved harmonic Poisson segmented regression model, and used the three models to explain the intervention effect, respectively. Finally, we selected the optimal model by comparing the fitting effects of the three models.

Results

The primary analysis showed an underlying upward trend of Kala-Azar before intervention [incidence rate ratio (IRR): 1.045, 95% confidence interval (CI): 1.027-1.063, p < 0.001]. In terms of long-term effects, the rise of Kala-Azar slowed down significantly after the intervention (IRR:0.960, 95%CI:0.927-0.995, p = 0.026), and the risk of Kala-Azar increased by 0.3% for each additional month after intervention (β1  + β3  = 0.003, IRR = 1.003). The results of the model fitting effect showed that the improved harmonic Poisson segmental regression model had the best fitting effect, and the values of MSE, MAE, and RMSE were the lowest, which were 0.017, 0.101, and 0.130, respectively.

Conclusion

In the long term, the intervention measures taken by the Yangquan CDC can well curb the upward trend of Kala-Azar. The improved harmonic Poisson segmented regression model has higher fitting performance, which can provide a certain scientific reference for the evaluation of the intervention effect of seasonal infectious diseases.

Is the expansion of the subway network alone capable of improving local air quality? A study case in São Paulo, Brazil

One of the policies adopted to reduce vehicular emissions is subway network expansion. This work fitted interrupted regression models to investigate the effects of the inauguration of subway stations on the mean, trend, and seasonality of the NO, NO2, NOx, and PM10 local concentrations. The regions investigated in the city of São Paulo (Brazil) were Pinheiros, Butantã, and St. Amaro. In Pinheiros, after the inauguration of the subway station, there were downward trends for all pollutants. However, these trends were not significantly different from the trends observed before. In Butantã, only regarding NO, there was a significant reduction and seasonal change after the subway station's inauguration. In St. Amaro, no trend in the PM10 concentration was noted. The absence of other transportation and land use policies in an integrative way to the subway network expansion may be responsible for the low air quality improvement. This study highlights that the expansion of the subway network must be integrated with other policies to improve local air quality.

Evaluating Montréal's harm reduction interventions for people who inject drugs: protocol for observational study and cost-effectiveness analysis

INTRODUCTION

The main harm reduction interventions for people who inject drugs (PWID) are supervised injection facilities, needle and syringe programmes and opioid agonist treatment. Current evidence supporting their implementation and operation underestimates their usefulness by excluding skin, soft tissue and vascular infections (SSTVIs) and anoxic/toxicity-related brain injury from cost-effectiveness analyses (CEA). Our goal is to conduct a comprehensive CEA of harm reduction interventions in a setting with a large, dispersed, heterogeneous population of PWID, and include prevention of SSTVIs and anoxic/toxicity-related brain injury as measures of benefit in addition to HIV, hepatitis C and overdose morbidity and mortalities averted.

METHODS AND ANALYSIS

This protocol describes how we will develop an open, retrospective cohort of adult PWID living in Québec between 1 January 2009 and 31 December 2020 using administrative health record data. By complementing this data with non-linkable paramedic dispatch records, regional monthly needle and syringe dispensation counts and repeated cross-sectional biobehavioural surveys, we will estimate the hazards of occurrence and the impact of Montréal's harm reduction interventions on the incidence of drug-use-related injuries, infections and deaths. We will synthesise results from our empirical analyses with published evidence to simulate infections and injuries in a hypothetical population of PWID in Montréal under different intervention scenarios including current levels of use and scale-up, and assess the cost-effectiveness of each intervention from the public healthcare payer's perspective.

ETHICS AND DISSEMINATION

This study was approved by McGill University's Institutional Review Board (Study Number: A08-E53-19B). We will work with community partners to disseminate results to the public and scientific community via scientific conferences, a publicly accessible report, op-ed articles and open access peer-reviewed journals.

Importance of applying Mixed Generalized Additive Model (MGAM) as a method for assessing the environmental health impacts: Ambient temperature and Acute Myocardial Infarction (AMI), among elderly in Shanghai, China

Association between acute myocardial infarction (AMI) morbidity and ambient temperature has been examined with generalized linear model (GLM) or generalized additive model (GAM). However, the effect size by these two methods might be biased due to the autocorrelation of time series data and arbitrary selection of degree of freedom of natural cubic splines. The present study analyzed how the climatic factors affected AMI morbidity for older adults in Shanghai with Mixed generalized additive model (MGAM) that addressed these shortcomings mentioned. Autoregressive random effect was used to model the relationship between AMI and temperature, PM10, week days and time. The degree of freedom of time was chosen based on the seasonal pattern of temperature. The performance of MGAM was compared with GAM on autocorrelation function (ACF), partial autocorrelation function (PACF) and goodness of fit. One-year predictions of AMI counts in 2011 were conducted using MGAM with the moving average. Between 2007 and 2011, MGAM adjusted the autocorrelation of AMI time series and captured the seasonal pattern after choosing the degree of freedom of time at 5. Using MGAM, results were well fitted with data in terms of both internal (R2 = 0.86) and external validity (correlation coefficient = 0.85). The risk of AMI was relatively high in low temperature (Risk ratio = 0.988 (95% CI 0.984, 0.993) for under 12°C) and decreased as temperature increased and speeded up within the temperature zone from 12°C to 26°C (Risk ratio = 0.975 (95% CI 0.971, 0.979), but it become increasing again when it is 26°C although not significantly (Risk ratio = 0.999 (95% CI 0.986, 1.012). MGAM is more appropriate than GAM in the scenario of response variable with autocorrelation and predictors with seasonal variation. The risk of AMI was comparatively higher when temperature was lower than 12°C in Shanghai as a typical representative location of subtropical climate.

A forecast model for prevention of foodborne outbreaks of non-typhoidal salmonellosis

Background

This work presents a forecast model for non-typhoidal salmonellosis outbreaks.

Method

This forecast model is based on fitted values of multivariate regression time series that consider diagnosis and estimation of different parameters, through a very flexible statistical treatment called generalized auto-regressive and moving average models (GSARIMA).

Results

The forecast model was validated by analyzing the cases of Salmonella enterica serovar Enteritidis in Sydney Australia (2014–2016), the environmental conditions and the consumption of high-risk food as predictive variables.

Conclusions

The prediction of cases of Salmonella enterica serovar Enteritidis infections are included in a forecast model based on fitted values of time series modeled by GSARIMA, for an early alert of future outbreaks caused by this pathogen, and associated to high-risk food. In this context, the decision makers in the epidemiology field can led to preventive actions using the proposed model.

Evaluating the intended and unintended consequences of opioid-prescribing interventions on primary care in British Columbia, Canada: protocol for a retrospective population-based cohort study

INTRODUCTION

Between 2015 and 2018, there were over 40 000 opioid-related overdose events and 4551 deaths among residents in British Columbia (BC). During this time the province mobilised a variety of policy levers to encourage physicians to expand access to opioid agonist treatment and the College of Physicians and Surgeons of British Columbia (CPSBC) released a practice standard establishing legally enforceable minimum thresholds of professional behaviour in the hopes of curtailing overdose events. Our goal is to conduct a comprehensive investigation of the intended and unintended consequences of these policy changes. Specifically, we aim to understand the effects of these measures on physician prescribing behaviours, identify physician characteristics associated with uptake of the new measures, and measure the effects of the policy changes on patients' access to quality primary care.

METHODS AND ANALYSIS

This is a population-level, retrospective cohort study of all BC primary care physicians who prescribed any opioid medication for opioid-use disorder or chronic non-cancer pain during the study period, and their patients. The study period is 1 January 2013-31 December 2018, with a 1-year wash-in period (1 January 2012-31 December 2012) to exclude patients who initiated long-term opioid treatment prior to our study period or whose pain type (ie, 'chronic non-cancer', 'acute', 'cancer or palliative', or 'other') cannot be confirmed. The project combines five administrative health datasets under the authority of the BC Ministry of Health, with the CPSBC's Physician Registry, BC Cancer Agency's Cancer Registry and Vital Statistics' Mortality data. We will create measures of prescribing concordance, access, continuity, and comprehensiveness to assess primary care delivery and quality at both the physician and patient level. We will use generalised estimating equations, interrupted time series, mixed effects models, and funnel plots to identify factors related to changes in prescribing and evaluate the impact of the changes to prescribing policies. Results will be reported using appropriate Enhancing the QUAlity and Transparency Of health Research guidelines (eg, STrengthening the Reporting of OBservational studies in Epidemiology).

ETHICS AND DISSEMINATION

This study has been approved by McGill University's Institutional Review Board (#A11-M55-19A), and the University of British Columbia's Research Ethics Board (#H19-03537). We will disseminate results via a combination of open access peer-reviewed journal publications, conferences, lay summaries and OpEds.

Modelling malaria reduction in a highly endemic country: Evidence from household survey, climate, and program data in Zambia

Substantial efforts have seen the reduction in malaria prevalence from 33% in 2006 to 19.4% in 2015 in Zambia. Many studies have used effect measures, such as odds ratios, of malaria interventions without combining this information with coverage levels of the interventions to assess how malaria prevalence would change if these interventions were scaled up. We contribute to filling this gap by combining intervention coverage information with marginal predictions to model the extent to which key interventions can bring down malaria in Zambia. We used logistic regression models and derived marginal effects using repeated cross-sectional survey data from the Malaria Indicator Survey (MIS) datasets for Zambia collected in 2010, 2012 and 2015. Average monthly temperature and rainfall data were obtained from climate explorer a satellite-generated database. We then conducted a counterfactual analysis using the estimated marginal effects and various hypothetical levels of intervention coverage to assess how different levels of coverage would affect malaria prevalence. Increasing IRS and ITNs from the 2015 levels of coverage of 28.9% and 58.9% respectively to at least 80% and rising standard housing to 20% from the 13.4% in 2015 may bring malaria prevalence down to below 15%. If the percentage of modern houses were increased further to 90%, malaria prevalence might decrease to 10%. Other than ITN and IRS, streamlining and increasing of the percentage of standard houses in malaria fight would augment and bring malaria down to the levels needed for focal malaria elimination. The effects of ITNs, IRS and Standard housing were pronounced in high than low epidemiological areas.

The study on the early warning period of varicella outbreaks based on logistic differential equation model

Chickenpox is a common acute and highly contagious disease in childhood; moreover, there is currently no targeted treatment. Carrying out an early warning on chickenpox plays an important role in taking targeted measures in advance as well as preventing the outbreak of the disease. In recent years, the infectious disease dynamic model has been widely used in the research of various infectious diseases. The logistic differential equation model can well demonstrate the epidemic characteristics of epidemic outbreaks, gives the point at which the early epidemic rate changes from slow to fast. Therefore, our study aims to use the logistic differential equation model to explore the epidemic characteristics and early-warning time of varicella. Meanwhile, the data of varicella cases were collected from first week of 2008 to 52nd week of 2017 in Changsha. Finally, our study found that the logistic model can be well fitted with varicella data, besides the model illustrated that there are two peaks of varicella at each year in Changsha City. One is the peak in summer–autumn corresponding to the 8th–38th week; the other is in winter–spring corresponding to the time from the 38th to the seventh week next year. The ‘epidemic acceleration week’ average value of summer–autumn and winter–spring are about the 16th week (ranging from the 15th to 17th week) and 45th week (ranging from the 44th to 47th week), respectively. What is more, taking warning measures during the acceleration week, the preventive effect will be delayed; thus, we recommend intervene during recommended warning weeks which are the 15th and 44th weeks instead.

A Dynamical and Zero-Inflated Negative Binomial Regression Modelling of Malaria Incidence in Limpopo Province, South Africa

Recent studies have considered the connections between malaria incidence and climate variables using mathematical and statistical models. Some of the statistical models focused on time series approach based on Box-Jenkins methodology or on dynamic model. The latter approach allows for covariates different from its original lagged values, while the Box-Jenkins does not. In real situations, malaria incidence counts may turn up with many zero terms in the time series. Fitting time series model based on the Box-Jenkins approach and ARIMA may be spurious. In this study, a zero-inflated negative binomial regression model was formulated for fitting malaria incidence in Mopani and Vhembe-two of the epidemic district municipalities in Limpopo, South Africa. In particular, a zero-inflated negative binomial regression model was formulated for daily malaria counts as a function of some climate variables, with the aim of identifying the model that best predicts reported malaria cases. Results from this study show that daily rainfall amount and the average temperature at various lags have a significant influence on malaria incidence in the study areas. The significance of zero inflation on the malaria count was examined using the Vuong test and the result shows that zero-inflated negative binomial regression model fits the data better. A dynamical climate-based model was further used to investigate the population dynamics of mosquitoes over the two regions. Findings highlight the significant roles of Anopheles arabiensis on malaria transmission over the regions and suggest that vector control activities should be intense to eradicate malaria in Mopani and Vhembe districts. Although An. arabiensis has been identified as the major vector over these regions, our findings further suggest the presence of additional vectors transmitting malaria in the study regions. The findings from this study offer insight into climate-malaria incidence linkages over Limpopo province of South Africa.

Application of a hybrid model in predicting the incidence of tuberculosis in a Chinese population

Objective: To investigate suitable forecasting models for tuberculosis (TB) in a Chinese population by comparing the predictive value of the autoregressive integrated moving average (ARIMA) model and the ARIMA-generalized regression neural network (GRNN) hybrid model. Methods: We used the monthly incidence rate of TB in Lianyungang city from January 2007 through June 2016 to construct a fitting model, and we used the incidence rate from July 2016 to December 2016 to evaluate the forecasting accuracy. The root mean square error (RMSE), mean absolute percentage error (MAPE), mean absolute error (MAE) and mean error rate (MER) were used to assess the performance of these models in fitting and forecasting the incidence of TB. Results: The ARIMA (10, 1, 0) (0, 1, 1)₁₂ model was selected from plausible ARIMA models, and the optimal spread value of the ARIMA-GRNN hybrid model was 0.23. For the fitting dataset, the RMSE, MAPE, MAE and MER were 0.5594, 11.5000, 0.4202 and 0.1132, respectively, for the ARIMA (10, 1, 0) (0, 1, 1)₁₂ model, and 0.5259, 11.2181, 0.3992 and 0.1075, respectively, for the ARIMA-GRNN hybrid model. For the forecasting dataset, the RMSE, MAPE, MAE and MER were 0.2805, 8.8797, 0.2261 and 0.0851, respectively, for the ARIMA (10, 1, 0) (0, 1, 1)₁₂ model, and 0.2553, 5.7222, 0.1519 and 0.0571, respectively, for the ARIMA-GRNN hybrid model. Conclusions: The ARIMA-GRNN hybrid model was shown to be superior to the single ARIMA model in predicting the short-term TB incidence in the Chinese population, especially in fitting and forecasting the peak and trough incidence.

A systematic review of aberration detection algorithms used in public health surveillance

The algorithms used for detecting anomalies have evolved substantially over the last decade to take advantage of advances in informatics and to accommodate changes in surveillance data. We identified 145 studies since 2007 that evaluated statistical methods used to detect aberrations in public health surveillance data. For each study, we classified the analytic methods and reviewed the evaluation metrics. We also summarized the practical usage of the detection algorithms in public health surveillance systems worldwide. Traditional methods (e.g., control charts, linear regressions) were the focus of most evaluation studies and continue to be used commonly in practice. There was, however, an increase in the number of studies using forecasting methods and studies applying machine learning methods, hidden Markov models, and Bayesian framework to multivariate datasets. Evaluation studies demonstrated improved accuracy with more sophisticated methods, but these methods do not appear to be used widely in public health practice.

Natural selection and the predictability of evolution in Timema stick insects

Predicting evolution remains difficult. We studied the evolution of cryptic body coloration and pattern in a stick insect using 25 years of field data, experiments, and genomics. We found that evolution is more difficult to predict when it involves a balance between multiple selective factors and uncertainty in environmental conditions than when it involves feedback loops that cause consistent back-and-forth fluctuations. Specifically, changes in color-morph frequencies are modestly predictable through time (r² = 0.14) and driven by complex selective regimes and yearly fluctuations in climate. In contrast, temporal changes in pattern-morph frequencies are highly predictable due to negative frequency-dependent selection (r² = 0.86). For both traits, however, natural selection drives evolution around a dynamic equilibrium, providing some predictability to the process.

Time series analysis based on two-part models for excessive zero count data to detect farm-level outbreaks of swine echinococcosis during meat inspections

Echinococcus multilocularis is a parasite that causes highly pathogenic zoonoses and is maintained in foxes and rodents on Hokkaido Island, Japan. Detection of E. multilocularis infections in swine is epidemiologically important. In Hokkaido, administrative information is provided to swine producers based on the results of meat inspections. However, as the current criteria for providing administrative information often results in delays in providing information to producers, novel criteria are needed. Time series models were developed to monitor autocorrelations between data and lags using data collected from 84 producers at the Higashi-Mokoto Meat Inspection Center between April 2003 and November 2015. The two criteria were quantitatively compared using the sign test for the ability to rapidly detect farm-level outbreaks. Overall, the time series models based on an autoexponentially regressed zero-inflated negative binomial distribution with 60th percentile cumulative distribution function of the model detected outbreaks earlier more frequently than the current criteria (90.5%, 276/305, p<0.001). Our results show that a two-part model with autoexponential regression can adequately deal with data involving an excessive number of zeros and that the novel criteria overcome disadvantages of the current criteria to provide an earlier indication of increases in the rate of echinococcosis.

Pneumococcal conjugate vaccines and hospitalization of children for pneumonia: a time-series analysis, South Africa, 2006-2014

OBJECTIVE

To assess the impact of immunization with pneumococcal conjugate vaccines on all-cause pneumonia hospitalizations among children in Soweto, South Africa.

METHODS

We used data collected at the Chris Hani Baragwanath Hospital in Soweto between 2006 and 2014 - i.e. before and after April 2009, when a pneumococcal conjugate vaccine was first included in South Africa's routine immunization programme. Using a Bayesian generalized seasonal autoregressive moving-average model and the data collected in 2006-2008, we estimated the numbers of children that would have been hospitalized for pneumonia between 2010 and 2014 if no pneumococcal conjugate vaccines had been used. These estimates were then compared with the corresponding numbers of hospitalizations observed.

FINDINGS

Between 2006 and 2014, 26 778 children younger than five years - including 3388 known to be infected with human immunodeficiency virus (HIV) - were admitted to the study hospital for pneumonia. We estimated that, for the children known to be infected with HIV and for the other children, pneumococcal conjugate vaccines reduced the numbers of hospitalizations for pneumonia in 2014 by 33% (50% credible interval, CrI: 6 to 52) and 39% (50% CrI: 24 to 50), respectively. In the study hospital in 2012-2014, as a result of immunizations with these vaccines, there were an estimated 3100 fewer pneumonia hospitalizations of children younger than five years.

CONCLUSION

In our study hospital, following the introduction of pneumococcal conjugate vaccines into the national immunization programme, there were significant reductions in pneumonia hospitalizations among children.

CUSUM chart to monitor autocorrelated counts using Negative Binomial GARMA model

Cumulative sum control charts have been used for health surveillance due to its efficiency to detect soon small shifts in the monitored series. However, these charts may fail when data are autocorrelated. An alternative procedure is to build a control chart based on the residuals after fitting autoregressive moving average models, but these models usually assume Gaussian distribution for the residuals. In practical health surveillance, count series can be modeled by Poisson or Negative Binomial regression, this last to control overdispersion. To include serial correlations, generalized autoregressive moving average models are proposed. The main contribution of the current article is to measure the impact, in terms of average run length on the performance of cumulative sum charts when the serial correlation is neglected in the regression model. Different statistics based on transformations, the deviance residual, and the likelihood ratio are used to build cumulative sum control charts to monitor counts with time varying means, including trend and seasonal effects. The monitoring of the weekly number of hospital admissions due to respiratory diseases for people aged over 65 years in the city São Paulo-Brazil is considered as an illustration of the current method.

Climate, environmental and socio-economic change: weighing up the balance in vector-borne disease transmission

Arguably one of the most important effects of climate change is the potential impact on human health. While this is likely to take many forms, the implications for future transmission of vector-borne diseases (VBDs), given their ongoing contribution to global disease burden, are both extremely important and highly uncertain. In part, this is owing not only to data limitations and methodological challenges when integrating climate-driven VBD models and climate change projections, but also, perhaps most crucially, to the multitude of epidemiological, ecological and socio-economic factors that drive VBD transmission, and this complexity has generated considerable debate over the past 10-15 years. In this review, we seek to elucidate current knowledge around this topic, identify key themes and uncertainties, evaluate ongoing challenges and open research questions and, crucially, offer some solutions for the field. Although many of these challenges are ubiquitous across multiple VBDs, more specific issues also arise in different vector-pathogen systems.

Malaria incidence from 2005-2013 and its associations with meteorological factors in Guangdong, China

BACKGROUND

The temporal variation of malaria incidence has been linked to meteorological factors in many studies, but key factors observed and corresponding effect estimates were not consistent. Furthermore, the potential effect modification by individual characteristics is not well documented. This study intends to examine the delayed effects of meteorological factors and the sub-population's susceptibility in Guangdong, China.

METHODS

The Granger causality Wald test and Spearman correlation analysis were employed to select climatic variables influencing malaria. The distributed lag non-linear model (DLNM) was used to estimate the non-linear and delayed effects of weekly temperature, duration of sunshine, and precipitation on the weekly number of malaria cases after controlling for other confounders. Stratified analyses were conducted to identify the sub-population's susceptibility to meteorological effects by malaria type, gender, and age group.

RESULTS

An incidence rate of 1.1 cases per 1,000,000 people was detected in Guangdong from 2005-2013. High temperature was associated with an observed increase in malaria incidence, with the effect lasting for four weeks and a maximum relative risk (RR) of 1.57 (95% confidence interval (CI): 1.06-2.33) by comparing 30°C to the median temperature. The effect of sunshine duration peaked at lag five and the maximum RR was 1.36 (95% CI: 1.08-1.72) by comparing 24 hours/week to 0 hours/week. A J-shaped relationship was found between malaria incidence and precipitation with a threshold of 150 mm/week. Over the threshold, precipitation increased malaria incidence after four weeks with the effect lasting for 15 weeks, and the maximum RR of 1.55 (95% CI: 1.18-2.03) occurring at lag eight by comparing 225 mm/week to 0 mm/week. Plasmodium falciparum was more sensitive to temperature and precipitation than Plasmodium vivax. Females had a higher susceptibility to the effects of sunshine and precipitation, and children and the elderly were more sensitive to the change of temperature, sunshine duration, and precipitation.

CONCLUSION

Temperature, duration of sunshine and precipitation played important roles in malaria incidence with effects delayed and varied across lags. Climatic effects were distinct among sub-groups. This study provided helpful information for predicting malaria incidence and developing the future warning system.

Remotely-sensed, nocturnal, dew point correlates with malaria transmission in Southern Province, Zambia: a time-series study

BACKGROUND

Plasmodium falciparum transmission has decreased significantly in Zambia in the last decade. The malaria transmission is influenced by environmental variables. Incorporation of environmental variables in models of malaria transmission likely improves model fit and predicts probable trends in malaria disease. This work is based on the hypothesis that remotely-sensed environmental factors, including nocturnal dew point, are associated with malaria transmission and sustain foci of transmission during the low transmission season in the Southern Province of Zambia.

METHODS

Thirty-eight rural health centres in Southern Province, Zambia were divided into three zones based on transmission patterns. Correlations between weekly malaria cases and remotely-sensed nocturnal dew point, nocturnal land surface temperature as well as vegetation indices and rainfall were evaluated in time-series analyses from 2012 week 19 to 2013 week 36. Zonal as well as clinic-based, multivariate, autoregressive, integrated, moving average (ARIMAX) models implementing environmental variables were developed to model transmission in 2011 week 19 to 2012 week 18 and forecast transmission in 2013 week 37 to week 41.

RESULTS

During the dry, low transmission season significantly higher vegetation indices, nocturnal land surface temperature and nocturnal dew point were associated with the areas of higher transmission. Environmental variables improved ARIMAX models. Dew point and normalized differentiated vegetation index were significant predictors and improved all zonal transmission models. In the high-transmission zone, this was also seen for land surface temperature. Clinic models were improved by adding dew point and land surface temperature as well as normalized differentiated vegetation index. The mean average error of prediction for ARIMAX models ranged from 0.7 to 33.5%. Forecasts of malaria incidence were valid for three out of five rural health centres; however, with poor results at the zonal level.

CONCLUSIONS

In this study, the fit of ARIMAX models improves when environmental variables are included. There is a significant association of remotely-sensed nocturnal dew point with malaria transmission. Interestingly, dew point might be one of the factors sustaining malaria transmission in areas of general aridity during the dry season.

Models for short term malaria prediction in Sri Lanka

Background

Malaria in Sri Lanka is unstable and fluctuates in intensity both spatially and temporally. Although the case counts are dwindling at present, given the past history of resurgence of outbreaks despite effective control measures, the control programmes have to stay prepared. The availability of long time series of monitored/diagnosed malaria cases allows for the study of forecasting models, with an aim to developing a forecasting system which could assist in the efficient allocation of resources for malaria control.

Methods

Exponentially weighted moving average models, autoregressive integrated moving average (ARIMA) models with seasonal components, and seasonal multiplicative autoregressive integrated moving average (SARIMA) models were compared on monthly time series of district malaria cases for their ability to predict the number of malaria cases one to four months ahead. The addition of covariates such as the number of malaria cases in neighbouring districts or rainfall were assessed for their ability to improve prediction of selected (seasonal) ARIMA models.

Results

The best model for forecasting and the forecasting error varied strongly among the districts. The addition of rainfall as a covariate improved prediction of selected (seasonal) ARIMA models modestly in some districts but worsened prediction in other districts. Improvement by adding rainfall was more frequent at larger forecasting horizons.

Conclusion

Heterogeneity of patterns of malaria in Sri Lanka requires regionally specific prediction models. Prediction error was large at a minimum of 22% (for one of the districts) for one month ahead predictions. The modest improvement made in short term prediction by adding rainfall as a covariate to these prediction models may not be sufficient to merit investing in a forecasting system for which rainfall data are routinely processed.