Study protocol

Study protocol: a systematic review and meta-analysis on the association of daily mean temperature and mortality in China

Updated 22 October 2018

Qianlai Luo, Shanshan Li, Yuming Guo, Xuemei Han, Jouni J. K. Jaakkola


1. Introduction

2. General and specific study objectives

3. Identifying relevant studies

4. Eligibility criteria for inclusion and exclusion of studies

5. Abstracting the data

6. Author contacts and additional information

7. Statistical analyses

8. Estimating study quality and publication bias

9. Recordkeeping and reporting

1. Introduction

There is substantial evidence that mortality increases in non-optimal temperatures in western countries, but less scientific evidence in other parts of the world. China, a country that consists one fifth of the world population, has seen a significant amount of new information on this topic published in recent years. In on our preliminary literature search, however, we only identified a scientific summary that was in Chinese language, published in a Chinese journal, and included 9 studies published until 2012 covering 15 cities in China. Preliminary literature search yielded over 100 articles. There seems to be heterogeneity in study designs, methods and measures used, and reporting practices.

We aim to summarize the evidence on hot and cold effects on mortality in China, including literature published in Chinese language, and share it with a wider scientific community. Due to the diversity described above, we anticipate that conducting a systematic review and meta-analysis on the topic is likely to differ from conducting a systematic review and meta-analysis on a clinical topic adhering to standard study designs and reported in standard epidemiological formats. This needs to be taken into account in the study protocol. To begin with, we anticipate the process of literature search, review and data extraction to be a reiterative process.

Here we create strictly a priori explicit procedures to reduce bias and ensure reproducibility of the study, while leaving certain pre-defined procedures flexibility needed for sensible synthesis. Our strategy is to maximise the amount of comparable information by utilising a systematic literature search of a wide spectrum, and include all articles in which Relative Risks associated with 1 degree increase or decrease from the reference points were available, or could be obtained through conversion.  

2. General and specific study objectives

Our general study objectives are to:

1) Summarise the evidence on the relation between hot/cold temperature and mortality in China;

2) Assess the pattern of associations across causes of deaths, socio-economic factors, and across climatic regions;

3) Identify possible gaps in knowledge.

Our specific study questions are:

  1. Is there an association between hot, cold temperature and mortality (including all-cause, and cause-specific mortality, age, sex, education status-specific) in China?

2) What is the strength and direction of these associations?

3) Is there a pattern of associations across different climatic region? 

4) What, if any, are the modifying factors?

3. Identifying relevant studies

We will conduct a systematic search of four databases [PubMed MEDLINE, SciVerse Scopus, Web of Science, and China Knowledge Resource Integrated Database (CNKI:]. Pubmed, Scopus and WoS are English databases; CNKI is a Chinese article database. 

English databases:

For the three English databases, we will construct tailored searches, using boolean logic “AND” to connect four above text strings indicating exposure, outcome, study type/methods and geographical coverage respectively.

 1) Exposure: "extreme weather", temperature, "extreme weather events", "climate change", "extreme temperature", cold, heat, "cold temperature", "hot temperature", "cold spell", “cold wave” “cold surge” "heat wave";

[temperature OR "extreme weather" OR "extreme weather events" OR "climate change" OR "extreme temperature" OR cold OR heat OR "cold temperature" OR "hot temperature" OR "cold spell" OR "heat wave"]

2) Outcome: mortality, death, still birth;

[mortality OR death OR stillbirth OR “still-birth” OR “still birth” ]

3) Study Type/Methods: "time series", "time-series OR time series", "case-crossover", cohort/DLN/reviews

["time series" OR “time-series” OR "case-crossover" OR “distributed lag non-linear” OR “distributed lag” OR “generalized additive model” OR cohort OR epidemiological OR association OR associated OR related OR relationship OR “systematic review” OR “meta-analysis” OR "literature review" OR “systematic literature review” ]

4) Geographical coverage: provinces in grey, name of cities in blue, added “” to city names that had ’ or space. China, Chinese, Taiwan, “Hong Kong”, “river delta”, and list of province names, and list of cities

[multi-country OR multicountry OR Cross-Country OR China OR Chinese OR “river delta” OR Beijing OR  Tianjin OR  Hebei OR  Shanxi OR  "Inner Mongolia" OR  Liaoning OR  Jilin OR  Heilongjiang OR  Shanghai OR  Jiangsu OR  Zhejiang OR  Anhui OR  Fujian OR  Jiangxi OR  Shandong OR  Henan OR  Hubei OR  Hunan OR  Guangdong OR  Guangxi OR  Hainan OR  Chongqing OR  Sichuan OR  Guizhou OR  Yunnan OR  Tibet OR  Shaanxi OR  Gansu OR  Qinghai OR  Ningxia OR  Xinjiang OR  "Hong Kong" OR  Macau OR Taiwan OR Karamay OR Lhasa OR Altay OR Ezhou OR Hegang OR Zhoushan OR Panzhihua OR Lijiang OR Laiwu OR “Ma'anshan” OR Zhanjiang OR Panjin OR Beihai OR Zhuhai OR Benxi OR Wuwei OR Fuxin OR Liaoyang OR Hezhou OR Yinchuan OR Linyi OR Shijiazhuang OR Nanyang OR Shenzhen OR Suzhou OR Harbin OR Baoding OR Guangzhou OR Tianjin OR Chengdu OR Beijing OR Dongying OR Haikou OR Huaibei OR Lishui OR Fushun OR Xining OR Huainan OR Aksu OR Yangjiang OR Yingkou OR Huangshi OR Dandong OR Jiamusi OR Huludao OR Nanping OR Baotou OR Chaozhou OR Zigong OR Tieling OR Xiangtan OR Putian OR Mudanjiang OR Rizhao OR Weihai OR Hohhot OR Daqing OR Qinhuangdao OR Shanghai OR Chongqing OR Ürümqi OR Zhenjiang OR Jinzhou OR Zhongshan OR Bengbu OR Tianshui OR Bazhong OR Datong OR Yulin OR Wuhu OR Jiaozuo OR Puyang OR Lanzhou OR Anshan OR Neijiang OR Baoji OR Zaozhuang OR Binzhou OR Liuzhou OR Zhuzhou OR Zhaoqing OR Chuzhou OR Kashgar OR Yichang OR Guigang OR Taiyuan OR Xuchang OR Linfen OR Guiyang OR Chifeng OR Langfang OR Lianyungang OR Jilin OR Jiangmen OR Yangzhou OR Yibin OR Jiaxing OR Zibo OR Changzhou OR Huizhou OR Mianyang OR Taizhou OR Kaifeng OR Suqian OR Jiujiang OR Guilin OR “Huai'an” OR Zhangzhou OR Bozhou OR Pingdingshan OR Shaoxing OR Nanchang OR Xianyang OR Xiamen OR Anyang OR Anqing OR Suzhou OR Jinhua OR Shantou OR Suihua OR Yueyang OR “Tai'an” OR Xiangyang OR Dezhou OR Qiqihar OR “Lu'an” OR Jingzhou OR Hefei OR Xinxiang OR Changde OR Liaocheng OR Maoming OR Qujing OR Taizhou OR Xinyang OR Zunyi OR Nanchong OR Wuxi OR Kunming OR Luoyang OR Nanning OR Dalian OR Jinan OR Yantai OR Changsha OR Shaoyang OR Xingtai OR Cangzhou OR Hengyang OR Foshan OR Fuzhou OR Yancheng OR Nantong OR Tangshan OR Fuyang OR Ningbo OR Changchun OR Nanjing OR Zhucheng OR Jining OR Shenyang OR Quanzhou OR Dongguan OR Heze OR “Xi'an” OR Xuzhou OR Zhengzhou OR Hangzhou OR Qingdao OR Zhoukou OR Weifang OR Wenzhou OR Handan OR Wuhan OR Haicheng OR Dingzhou OR Fuqing OR Ganzhou OR Zhangqiu OR Yiwu OR Zoucheng OR Huazhou OR Rugao OR Qidong OR Haimen OR Dengzhou OR Cixi OR Zhuji OR Zhangjiagang OR Changshu OR Feicheng OR Wenling OR Taixing OR Linhai OR Jingjiang OR Shouguang OR Yixing OR Jiangyin OR Zaoyang OR Yuzhou OR Xinyi OR Pizhou OR Tengzhou OR Wuchuan OR Danyang OR Xiangcheng]

3.1. PubMed MEDLINE
A systematic literature search will be conducted using the PubMed MEDLINE database (National Library of Medicine, Bethesda, MD, USA) for all years available until the day of the search. We will search terms from title/abstract. And we will not use the “human” filter since some recent publications may not have been assigned labels yet for filtering. 

3.2. SciVerse Scopus
A systematic literature search will be conducted using the SciVerse Scopus database (Elsevier B.V., NL, EU) for all years available until the day of the search. The Health Sciences sub-database will be searched with above search terms in Ti/Abs/Key (Article Title, Abstract, Keywords, all years). We will limit the results to the following subjects: Medicine, Environmental Science, and language to English and Chinese.

3.3. Web of Science
A systematic literature search will be conducted using the Web of Science database (Thomson Reuters, NY, USA) for all years available until the day of the search. The 4 strings combined with “AND” will be searched in Topic field (all years, English and Chinese).

Chinese Database

3.4. China National Knowledge Infrastructure (CNKI)
After consulting the CNKI customer service, we decide to use the “fuzzy” search on CNKI, search will be Subject气温 (fuzzy)  AND Subject (死亡 OR死胎) , and will limit to 医药卫生科技 (medicine, hygiene science and technology). Choosing “Subject” will automatically include title, abstract and keywords.   

The reference lists of all studies included in the systematic review will be manually checked using the same eligibility criteria and steps of inclusion/exclusion.


4. Eligibility criteria for inclusion and exclusion of studies

Three eligibility criteria will be used to select studies for this systematic review. The study is deemed eligible if it:

a) is an original peer-reviewed study that had an independent study population; and

b) provides information on the relation between hot and cold temperature and human mortality in China; and

c) is either time-series or case-crossover study.

The articles on temperature and mortality that we found in our preliminary search had various ways of measuring temperature, reported multiple causes of deaths, and had various study populations and study settings.

For the current review and meta-analysis, we further apply the following inclusion criteria:

a) uses daily mean temperature as their temperature exposure;

b) causes of deaths are any of the following: all-cause/non-accidental, cardio-vascular, respiratory, and cerebro-vascular;

c) study population is whole population, rather than any specific high-risk group (e.g., studies that only included patients of a pre-existing disease were not included);

d) provides relative risk estimates on hot and cold effect.

We will exclude studies that looked only at out-of-hospital mortality, because we consider out-of-hospital mortality to be more concerned with healthcare delivery systems, which is not the focus of this study.

5. Abstracting the data

We will use the Cochrane data collection form ( to guide the design of a data collection form. The final contents of data extraction table will be decided after the systematic literature search has been completed and the contents of all studies have been reviewed.

Data will be sought for quantitative information on the health effects of hot and cold temperatures. Main characteristics of each included study will be recorded to a pilot-tested data extraction form. Studies shortlisted in our initial search will be independently reviewed by 2 investigators. Differences in decisions will be resolved through team discussion.

6. Author contacts and additional information

If specific information is unavailable to be extracted in the article, or are provided in an alternative format (e.g., presented in figures rather than in tables) in the shortlisted article, authors of these articles will be contacted to obtain these (original) data. Records will be kept of such correspondence.

7. Statistical analyses

We will use the standard methods of meta-analysis. We will conduct meta-analysis, and possibly meta-regression, depending on whether we have enough power. We will use Stata statistical package for meta-analysis.

Appropriate data processing measures might involve combining population groups such as age or education groups, combining multiple analyses. Effect estimates with 95% confidence intervals can be calculated using any suitable method or formulae available to maximize the amount of data in comparable format for the meta-analyses.

8. Estimating study quality and publication bias

Studies that lacks key data for the data extraction will not be included. Study quality of the included studies will be assessed using quality scale that suits the scope of the included study types. Publication bias will be assessed using funnel plots if the number of independent studies enables reliable use of the method.

9. Recordkeeping and reporting

Records will be kept of the progress of the study in accordance to good scientific practice. A separate record will be kept of the systematic literature searches, including the eligibility criteria applied to each entry. PRISMA guidelines will be followed in the reporting of this study.

Last updated: 22.10.2018