Enhancing estimation methods for integrating probability and nonprobability survey samples with machine-learning techniques. An application to a Survey on the impact of the COVID-19 pandemic in Spain.

Web surveys have replaced Face-to-Face and computer assisted telephone interviewing (CATI) as the main mode of data collection in most countries. This trend was reinforced as a consequence of COVID-19 pandemic-related restrictions. However, this mode still faces significant limitations in obtaining probability-based samples of the general population. For this reason, most web surveys rely on nonprobability survey designs. Whereas probability-based designs continue to be the gold standard in survey sampling, nonprobability web surveys may still prove useful in some situations. For instance, when small subpopulations are the group under study and probability sampling is unlikely to meet sample size requirements, complementing a small probability sample with a larger nonprobability one may improve the efficiency of the estimates. Nonprobability samples may also be designed as a mean for compensating for known biases in probability-based web survey samples by purposely targeting respondent profiles that tend to be underrepresented in these surveys. This is the case in the Survey on the impact of the COVID-19 pandemic in Spain (ESPACOV) that motivates this paper. In this paper, we propose a methodology for combining probability and nonprobability web-based survey samples with the help of machine-learning techniques. We then assess the efficiency of the resulting estimates by comparing them with other strategies that have been used before. Our simulation study and the application of the proposed estimation method to the second wave of the ESPACOV Survey allow us to conclude that this is the best option for reducing the biases observed in our data.

Platelet parameters and risk of hypertension disorders of pregnancy: a propensity score adjusted analysis.

While the role of platelets in cardiovascular diseases among the general population has been widely reported, evidence is inconsistent regarding the association between platelet indices with hypertension in pregnant women. In this study, we explored the associations between platelet parameters before 20 gestational weeks, an understudied period, with hypertensive disorders of pregnancy (HDP), including preeclampsia/eclampsia (PEEC) and gestational hypertension (GH). Based on the Born in Guangzhou Cohort Study, 12053 singleton pregnant women with platelet parameters, including platelet count (PC), mean platelet volume (MPV), plateletcrit (PCT), and platelet distribution width (PDW) measured at 14-19 gestational weeks were included. Conventional multivariable adjustment and propensity score analysis were used to control for confounders. The restricted cubic spline showed that the risk of PEEC increased linearly for PC, and non-linearly for PCT. For GH, the risk increased linearly for PC, MPV, and PCT, and non-linearly for PDW. When these indices were categorized into quintiles, women with higher PC and PCT were associated with increased risk of both PEEC and GH. Women with MPV exceeding the second quintile (≥ 8.8 fL) had a greater risk for GH, but not for PEEC. When HDP was classified into two groups (early- vs late-onset) based on the occurrence time, significant associations persisted for early-onset PEEC, early-onset GH, and late-onset GH. In conclusion, increased PC and PCT before 20 weeks of gestation were both associated with higher risk of PEEC and GH, while elevated MPV was only linked to GH.

Comparisons of breast conserving therapy versus mastectomy in young and old women with early-stage breast cancer: long-term results using propensity score adjustment method.

PURPOSE: This study aimed to compare the effect of BCT versus mastectomy on the recurrence and survival of different-aged patients, and to investigate whether effects of BCT versus mastectomy on survival of young patients were consistent with those of old patients. METHODS: Data on women with primary invasive breast cancer between 2007 and 2011 were extracted from the institutional database of Breast Center. Disparities in hormone receptor, tumor size, lymph node status, and Her-2 status between BCT and mastectomy groups were adjusted using the propensity score (PS) adjustment method. Patients were divided by age into two groups (≤ 40 years and > 40 years). We assessed proportions of local recurrence (LR), distant disease-free survival (DDFS), disease-free survival (DFS), and breast cancer-specific survival (BCSS) in different-aged groups; this assessment was further stratified by surgical treatment. RESULTS: A total of 2964 patients were included; 565 (19%) were aged ≤ 40 years. In the entire cohort, hazard ratios (HR) of BCT versus mastectomy for DDFS and DFS were 0.56 (P = 0.029) and 0.55 (P = 0.008), respectively. After PS adjustment, there was no significant difference between BCT and mastectomy in LR, DDFS, DFS and BCSS in the young age group. In the old age group, women who underwent BCT exhibited improved DDFS (HR 0.57, 95% CI 0.39-0.84, P = 0.004). CONCLUSIONS: BCT did not significantly affect survival outcomes of young patients with breast cancer. Superior survival of BCT compared to mastectomy was observed only in old patients.

An evaluation of whether propensity score adjustment can remove the self-selection bias inherent to web panel surveys addressing sensitive health behaviours.

BACKGROUND: In health research, population estimates are generally obtained from probability-based surveys. In market research surveys are frequently conducted from volunteer web panels. Propensity score adjustment (PSA) is often used at analysis to try to remove bias in the web survey, but empirical evidence of its effectiveness is mixed. We assess the ability of PSA to remove bias in the context of sensitive sexual health research and the potential of web panel surveys to replace or supplement probability surveys. METHODS: Four web panel surveys asked a subset of questions from the third British National Survey of Sexual Attitudes and Lifestyles (Natsal-3). Five propensity scores were generated for each web survey. The scores were developed from progressively larger sets of variables, beginning with demographic variables only and ending with demographic, sexual identity, lifestyle, attitudinal and sexual behaviour variables together. The surveys were weighted to match Natsal-3 based on propensity score quintiles. The performance of each survey and weighting was assessed by calculating the average 'absolute' odds ratio (inverse of the odds ratio if less than 1) across 22 pre-specified sexual behaviour outcomes of interest comparing the weighted web survey with Natsal-3. The average standard error across odds ratios was examined to assess the impact of weighting upon variance. RESULTS: Propensity weighting reduced bias relative to Natsal-3 as more variables were added for males, but had little effect for females, and variance increased for some surveys. Surveys with more biased estimates before propensity weighting showed greater reduction in bias from adjustment. Inconsistencies in performance were evident across surveys and outcomes. For most surveys and outcomes any reduction in bias was only partial and for some outcomes the bias increased. CONCLUSIONS: Even after propensity weighting using a rich range of information, including some sexual behaviour variables, some bias remained and variance increased for some web surveys. Whilst our findings support the use of PSA for web panel surveys, the reduction in bias is likely to be partial and unpredictable, consistent with the findings from market research. Our results do not support the use of volunteer web panels to generate unbiased population health estimates.

Peak C-reactive protein levels do not predict 30-day mortality for bacteremia: A retrospective cohort study.

INTRODUCTION: Although peak C-reactive protein (CRP) levels are correlated with the prognosis of some diseases, there have been no reports regarding the association between peak CRP levels and mortality in patients with bacteremia. The present study aimed to determine the association between peak CRP levels and prognosis in patients with bacteremia. METHODS: This retrospective cohort study was conducted in a single tertiary hospital and included patients with bacteremia admitted to the emergency department from November 2012 to March 2017. Cox regression analysis was performed to examine the association between peak CRP levels and 30-day mortality. We also performed propensity score adjustment using potential confounding factors. RESULTS: One hundred fifty-nine patients were included in the study. Peak CRP levels were significantly higher in the ß-hemolytic streptococci (P = 0.001) and Streptococcus pneumoniae (P = 0.003) groups. The C-statistic of the multivariate logistic regression model for the propensity score was 0.88. For 30-day mortality, peak CRP levels >20 mg/dL did not show significance in the Cox regression analysis (hazard ratio, 0.866; 95% confidence interval, 0.489-1.537; P = 0.62). Even after propensity score adjustment, no significance was noted (hazard ratio, 0.865; 95% confidence interval, 0.399-1.876; P = 0.71). CONCLUSIONS: Peak CRP levels were not an independent predictor of mortality in patients with bacteremia in the emergency department. Clinicians should consider that patients with extremely high peak CRP levels do not necessarily have high mortality and vice versa.

Outcome of thrombus aspiration in STEMI patients: a propensity score-adjusted study.

The use of thrombus aspiration (TA) prior to primary percutaneous coronary intervention (PPCI) has undergone a radical change in intervention guidelines. The clinical implications, however, are still under scrutiny. This study investigated the clinical effects and outcome of TA before PPCI in patients with ST-segment elevation myocardial infarction (STEMI). Overall 1027 patients with STEMI were analyzed in this retrospective, propensity score-adjusted, multicenter study. The primary endpoints were in-hospital and long-term mortality. There were 418 patients in the TA group and 609 in the conventional PPCI group. The in-hospital mortality rate was significantly higher in the TA group (8.7 vs. 5.0%; P = 0.03). During long-term follow-up [median follow-up duration 689 days (IQR 405-959)] the mortality rates were similar (TA 14.3%, conventional PPCI 15.0%; P = 0.85). Survival analysis for the complete observation period revealed no significant benefit of TA [hazard ratio (HR) 1.12; 97.5% CI 0.90-0.71; P = 0.63]. There were also no significant differences between the groups in the following secondary endpoints: composite of cardiovascular death and non-fatal reinfarction at discharge (P = 0.39), post-PPCI thrombolysis in myocardial infarction flow-grade-3 (P = 0.14), left ventricular ejection fraction (P = 0.47), and non-fatal reinfarction during follow-up (P = 0.17). Rehospitalization rate (1.82 vs. 10.3%; P < 0.0001) and Canadian Cardiovascular Society (CCS) grading (P = 0.02) during follow-up were significantly lower in the TA group. In our cohort the in-hospital mortality rate was significantly higher for TA patients, but during long-term follow-up the mortality rates did not differ. The incidence of rehospitalization and CCS grading were lower in the TA-treated patients.

Short-term outcomes of laparoscopic vs. open liver resection for hepatocellular adenoma: a multicenter propensity score adjustment analysis by the AFC-HCA-2013 study group.

BACKGROUND: Patients with hepatocellular adenomas are, in selected cases, candidates for liver resection, which can be approached via laparoscopy or laparotomy. The present study aimed to investigate the effects of the surgical approach on the postoperative morbidities of both minor and major liver resections. METHODS: In this multi-institutional study, all patients who underwent open or laparoscopic hepatectomies for hepatocellular adenomas between 1989 and 2013 in 27 European centers were retrospectively reviewed. A multiple imputation model was constructed to manage missing variables. Comparisons of both the overall rate and the types of complications between open and laparoscopic hepatectomy were performed after propensity score adjustment (via the standardized mortality ratio weighting method) on the factors that influenced the choice of the surgical approach. RESULTS: The laparoscopic approach was selected in 208 (38%) of the 533 included patients. There were 194 (93%) women. The median age was 38.9 years. After the application of multiple imputation, 208 patients who underwent laparoscopic operations were compared with 216 patients who underwent laparotomic operations. After adjustment, there were 20 (9.6%) major liver resections in the laparoscopy group and 17 (7.9%) in the open group. The conversion rate was 6.3%. The two surgical approaches exhibited similar postoperative morbidity rates and severities. Laparoscopic resection was associated with significantly less blood loss (93 vs. 196 ml, p < 0.001), a less frequent need for pedicle clamping (21 vs. 40%, p = 0.002), a reduced need for transfusion (8 vs. 24 red blood cells units, p < 0.001), and a shorter hospital stay (5 vs. 7 days, p < 0.001). The mortality was nil. CONCLUSIONS: Laparoscopy can achieve short-term outcomes similar to those of open surgery for hepatocellular adenomas and has the additional benefits of a reduced blood loss, need for transfusion, and a shorter hospital stay.

Annual surveillance mammography after early-stage breast cancer and breast cancer mortality.

BACKGROUND: After treatment for early-stage breast cancer (bca), annual surveillance mammography (asm) is recommended based on the assumption that early detection of an invasive ipsilateral breast tumour recurrence or subsequent invasive contralateral primary bca reduces bca mortality. METHODS: We studied women with unilateral early-stage bca treated by breast-conserving surgery from 1994 to 1997 who subsequently developed an ipsilateral recurrence or contralateral primary more than 24 months after initial diagnosis, without prior regional or distant metastases. Annual surveillance mammography was defined as 2 episodes of bilateral mammography 11-18 months apart during the 2 years preceding the ipsilateral recurrence or contralateral primary. The association between asm and bca death was evaluated using a Cox proportional hazards model. RESULTS: We identified 669 women who experienced invasive ipsilateral recurrence (n = 455) or a contralateral primary (n = 214) at a median interval of 53 months [interquartile range (iqr): 37-72 months] after initial diagnosis, 64.7% of whom had received asm during the preceding 2 years. The median interval between the 2 bilateral mammograms was 12.3 months (iqr: 11.9-13.0 months), and the median interval between the 2nd mammogram and histopathologic confirmation of ipsilateral recurrence or contralateral primary was 1.5 months (iqr: 0.8-3.9 months). Median followup after ipsilateral recurrence or contralateral primary was 7.76 years (iqr: 3.68-9.81 years). The adjusted hazard ratio for bca death associated with asm was 0.86 (95% confidence limits: 0.63, 1.16). CONCLUSIONS: Annual surveillance mammography was associated with a modestly lowered hazard ratio for bca death.

The effect of free health care on polypharmacy: a comparison of propensity score methods and multivariable regression to account for confounding.

PURPOSE: Differing healthcare access has implications for public health. In Ireland, eligibility for free public health care is means tested. Here, we examine the association between healthcare access and polypharmacy while accounting for underlying socio-economic and health status differences. METHODS: Self-reported regular medication use, history of diagnosed health conditions, disability, socio-demographics, and objective measures of depression and anxiety for adults aged 50-69 years (n = 5796) were ascertained from the population-representative Irish Longitudinal Study on Ageing. Objective measures of frailty, cognition, hypertension, and body mass index were also assessed for 4241 participants. The associations between free healthcare access and polypharmacy and use of 15 medication classes were estimated using multivariable modified Poisson regression, adjustment for the propensity score, and inverse probability of treatment weighting by the propensity score. RESULTS: Polypharmacy was reported by 22% and 7% of the 1932 and 3864 participants with and without public healthcare coverage. Public patients had a 21-38% greater risk of polypharmacy depending on the method used to account for confounding. Results were less robust using propensity score weighting. There was evidence that classes of cardiovascular drugs, drugs for acid-related disorders, and analgesics were used more commonly in public patients. Associations were mostly unaffected after also accounting for objective health measures but were significantly attenuated after accounting for frequency of healthcare visits. CONCLUSIONS: Publically funded health care in Ireland leads to greater medication use in people aged 50-69 years. This may reflect over-prescribing to public patients or restricted use among those who pay out of pocket.

Comparative effectiveness and safety of oral anticoagulants for atrial fibrillation: A retrospective cohort study.

Oral anticoagulants (OACs) are high-priority medications, frequently used with clinically important benefit and serious harm. Our objective was to compare the safety and effectiveness of direct-acting oral anticoagulants (DOACs) versus warfarin in a population where anticoagulation management and DOACs were readily available. A retrospective cohort study of all adults living in British Columbia with a diagnosis of atrial fibrillation and a first prescription for an OAC was conducted. Co-primary outcomes were ischemic stroke and systemic embolism, and major bleeding. Secondary outcomes included a net clinical outcome composite and analysis of discontinuation, switching, and key subgroups. We estimated the effects of treatment using time-to-event models with high-dimensional propensity score adjustment to control confounding. After adjustment for prescribing bias, a cohort (n = 20,113, 43.8% female, mean age 72.4 years) with a mean follow-up of 18.1 months showed that patients taking warfarin tended to be poorer, sicker, and less likely to have a cardiologist prescriber. Outcome event rates were not significantly different for DOACs compared to warfarin [adjusted rate ratio of 1.15 (0.91, 1.46) for systemic embolism, 0.94 (0.82, 1.08) for major bleeding, and 0.98 (0.90, 1.06) for net clinical outcome]. Only the effect of age on net clinical outcome met our strict criteria for predicting which group might be superior. Switch of drug class was associated with increased risk of events (p < 0.003). In this population, we found no difference in important clinical outcomes between warfarin and DOACs. Switching compared to not switching was associated with harm.

Overview and Initial Results of the National Center for Health Statistics' Research and Development Survey.

The National Center for Health Statistics is assessing the usefulness of recruited web panels in multiple research areas. One research area examines the use of close-ended probe questions and split-panel experiments for evaluating question-response patterns. Another research area is the development of statistical methodology to leverage the strength of national survey data to evaluate, and possibly improve, health estimates from recruited panels. Recruited web panels, with their lower cost and faster production cycle, in combination with established population health surveys, may be useful for some purposes for statistical agencies. Our initial results indicate that web survey data from a recruited panel can be used for question evaluation studies without affecting other survey content. However, the success of these data to provide estimates that align with those from large national surveys will depend on many factors, including further understanding of design features of the recruited panel (e.g. coverage and mode effects), the statistical methods and covariates used to obtain the original and adjusted weights, and the health outcomes of interest.

Adverse pregnancy and perinatal outcome in patients with recurrent pregnancy loss: Multiple imputation analyses with propensity score adjustment applied to a large-scale birth cohort of the Japan Environment and Children's Study.

PROBLEM: Several studies have reported the increased risk of preterm birth, premature rupture of membranes, and low birth weight in patients with recurrent pregnancy loss (RPL). There have been a limited number of large population-based studies examining adverse pregnancy and perinatal outcome after RPL. Multiple-imputed analyses (MIA) adjusting for biases due to missing data is also lacking. METHOD OF STUDY: A nationwide birth cohort study known as the "Japan Environment and Children's Study (JECS)" was conducted by the Ministry of the Environment. The subjects consisted of 104 102 registered children (including fetuses or embryos). RESULTS: No increased risk of a congenital anomaly, aneuploidy, neonatal asphyxia, or a small for date infant was observed among the children from women with a history of RPL. A novel increased risk of placental adhesion and uterine infection was found. The adjusted ORs using MIA in women with three or more PL were 1.76 (95% CI, 1.04-2.96) for a stillbirth, 1.68 (1.12-2.52) for a pregnancy loss, 2.53 (1.17-5.47) for placental adhesion, 1.87 (1.37-2.55) and 1.60 (.99-2.57) for mild and severe hypertensive disorders of pregnancy, respectively, 1.94 (1.06-3.55) for uterine infection, 1.28 (1.11-1.47) for caesarean section and .86 (.76-.98) for a male infant. CONCLUSION: MIA better quantified the risk, which could encourage women who might hesitate to attempt a subsequent pregnancy.

Retrospective Comparison of Single-Port Sleeve Gastrectomy Versus Three-Port Laparoscopic Sleeve Gastrectomy: a Propensity Score Adjustment Analysis.

PURPOSE: Evaluate the efficacy of single-port sleeve gastrectomy (SPSG) and then compare it to a less-invasive sleeve approach (three-port) (3PSG) according to a propensity score (PS) matching analysis. MATERIALS AND METHODS: We analyzed all patients who underwent SG through a three-port or a single-port laparoscopic approach. RESULTS: After 2 years, the follow-up was completed in 84% patients treated with 3PSG and 95% patients of the SPSG group. Excess weight loss (EWL) was comparable for the first year of follow-up within the two groups except for the controls at 3 months in which the SPSG group showed a higher EWL (p = 0.0243). CONCLUSION: We demonstrated the efficacy of SPSG in bariatric surgery even compared to another, less invasive, laparoscopic SG approach (three-port).

Measurement equivalence in mixed mode surveys.

Surveys increasingly use mixed mode data collection (e.g., combining face-to-face and web) because this controls costs and helps to maintain good response rates. However, a combination of different survey modes in one study, be it cross-sectional or longitudinal, can lead to different kinds of measurement errors. For example, respondents in a face-to-face survey or a web survey may interpret the same question differently, and might give a different answer, just because of the way the question is presented. This effect of survey mode on the question-answer process is called measurement mode effect. This study develops methodological and statistical tools to identify the existence and size of mode effects in a mixed mode survey. In addition, it assesses the size and importance of mode effects in measurement instruments using a specific mixed mode panel survey (Netherlands Kinship Panel Study). Most measurement instruments in the NKPS are multi-item scales, therefore confirmatory factor analysis (CFA) will be used as the main analysis tool, using propensity score methods to correct for selection effects. The results show that the NKPS scales by and large have measurement equivalence, but in most cases only partial measurement equivalence. Controlling for respondent differences on demographic variables, and on scale scores from the previous uni-mode measurement occasion, tends to improve measurement equivalence, but not for all scales. The discussion ends with a review of the implications of our results for analyses employing these scales.