Joint modelling of survival and backwards recurrence outcomes: an analysis of factors associated with fertility treatment in the U.S

The motivation for this paper is to determine factors associated with time-to-fertility treatment (TTFT) among women currently attempting pregnancy in a cross-sectional sample. Challenges arise due to dependence between time-to-pregnancy (TTP) and TTFT. We propose appending a marginal accelerated failure time model to identify risk factors of TTFT with a model for TTP where fertility treatment is included as a time-varying treatment to account for their dependence. The latter requires extending backwards recurrence survival methods to incorporate time-varying covariates with time-varying coefficients. Since backwards recurrence survival methods are a function of mean survival, computational difficulties arise in formulating mean survival when fertility treatment is unobserved, i.e. when TTFT is censored. We address these challenges by developing computationally friendly forms for the double expectation of TTP and TTFT. The performance is validated via comprehensive simulation studies. We apply our approach to the National Survey of Family Growth and explore factors related to prolonged TTFT in the U.S.

The quality of vital signs measurements and value preferences in electronic medical records varies by hospital, specialty, and patient demographics

We aimed to assess the frequency of value preferences in recording of vital signs in electronic healthcare records (EHRs) and associated patient and hospital factors. We used EHR data from Oxford University Hospitals, UK, between 01-January-2016 and 30-June-2019 and a maximum likelihood estimator to determine the prevalence of value preferences in measurements of systolic and diastolic blood pressure (SBP/DBP), heart rate (HR) (readings ending in zero), respiratory rate (multiples of 2 or 4), and temperature (readings of 36.0 °C). We used multivariable logistic regression to investigate associations between value preferences and patient age, sex, ethnicity, deprivation, comorbidities, calendar time, hour of day, days into admission, hospital, day of week and speciality. In 4,375,654 records from 135,173 patients, there was an excess of temperature readings of 36.0 °C above that expected from the underlying distribution that affected 11.3% (95% CI 10.6-12.1%) of measurements, i.e. these observations were likely inappropriately recorded as 36.0 °C instead of the true value. SBP, DBP and HR were rounded to the nearest 10 in 2.2% (1.4-2.8%) and 2.0% (1.3-5.1%) and 2.4% (1.7-3.1%) of measurements. RR was also more commonly recorded as multiples of 2. BP digit preference and an excess of temperature recordings of 36.0 °C were more common in older and male patients, as length of stay increased, following a previous normal set of vital signs and typically more common in medical vs. surgical specialities. Differences were seen between hospitals, however, digit preference reduced over calendar time. Vital signs may not always be accurately documented, and this may vary by patient groups and hospital settings. Allowances and adjustments may be needed in delivering care to patients and in observational analyses and predictive tools using these factors as outcomes or exposures.

Digit preference and biased conclusions in cardiac arrest studies

BACKGROUND

In cardiac arrest (CA), time is directly predictive of patients' prognosis. The increase in mortality resulting from delayed cardiopulmonary resuscitation has been quantified minute by minute. Times reported in CA management studies could reflect a timestamping bias referred to as "digit preference". This phenomenon leads to a preference for certain numerical values (such as 2, 5, or 10) over others (such as 13). Our objective was to investigate whether or not digit preference phenomenon could be observed in reported times of the day related to CA management, as noted in a national registry.

METHODS

We analyzed data from the French National Electronic Registry of Cardiac Arrests. We analyzed twelve times-of-the-day corresponding to each of the main steps of CA management reported by the emergency physicians who managed the patients in prehospital settings. We postulated that if CA occurred at random times throughout the day, then we could expect to see events related to CA management occurring at a similar rate each minute of each hour of the day, at a fraction of 1/60. We compared the fraction of times reported as multiples of 15 (0, 15, 30, and 45 - on the hour, quarters, half hour) with the expected fraction of 4/60 (i.e. 4 × 1/60).

MAIN RESULTS

A total of 47,211 times-of-the-day in relation to 6131 CA were analyzed. The most overrepresented numbers were: 0, with 3737 occurrences (8% vs 2% expected, p < 0.0001) and 30, with 2807 occurrences (6% vs 2% expected, p < 0.0001). Times-of-the-day as multiples of 15 were overrepresented (22% vs 7% expected, p < 0.0001).

CONCLUSION

Prospectively collected times were considerably influenced by digit preference phenomenon. Studies that are not based on automatic time recordings and that have not evaluated and considered this bias should be interpretated with caution.

Timing errors and temporal uncertainty in clinical databases-A narrative review

A firm concept of time is essential for establishing causality in a clinical setting. Review of critical incidents and generation of study hypotheses require a robust understanding of the sequence of events but conducting such work can be problematic when timestamps are recorded by independent and unsynchronized clocks. Most clinical models implicitly assume that timestamps have been measured accurately and precisely, but this custom will need to be re-evaluated if our algorithms and models are to make meaningful use of higher frequency physiological data sources. In this narrative review we explore factors that can result in timestamps being erroneously recorded in a clinical setting, with particular focus on systems that may be present in a critical care unit. We discuss how clocks, medical devices, data storage systems, algorithmic effects, human factors, and other external systems may affect the accuracy and precision of recorded timestamps. The concept of temporal uncertainty is introduced, and a holistic approach to timing accuracy, precision, and uncertainty is proposed. This quantitative approach to modeling temporal uncertainty provides a basis to achieve enhanced model generalizability and improved analytical outcomes.

Estimation of older-adult mortality from information distorted by systematic age misreporting

Testing theories about human senescence and longevity demands accurate information on older-adult mortality; this is rare in low- to middle-income countries where raw data may be distorted by defective completeness and systematic age misreporting. For this reason, such populations are frequently excluded from empirical tests of mortality and longevity theories, thus limiting their reach, as they reflect only a small and selected human mortality experience. In this paper we formulate an integrated method to compute estimates of older-adult mortality when vital registration and population counts are defective due to inaccurate coverage and/or systematic age misreporting. The procedure is validated with a simulation study that identifies a strategy to compute adjustments, which, under some assumptions, performs quite well. While the paper focuses on Latin American and Caribbean countries, the method is quite general and, with additional information and some model reformulation, could be applied to other populations with similar problems.

How to estimate mortality trends from grouped vital statistics

Background

Mortality data at the population level are often aggregated in age classes, for example 5-year age groups with an open-ended interval for the elderly aged 85+. Capturing detailed age-specific mortality patterns and mortality time trends from such coarsely grouped data can be problematic at older ages, especially where open-ended intervals are used.

Methods

We illustrate the penalized composite link model (PCLM) for ungrouping to model cancer mortality surfaces. Smooth age-specific distributions from data grouped in age classes of adjacent calendar years were estimated by constructing a two-dimensional regression, based on B-splines, and maximizing a penalized likelihood. We show the applicability of the proposed model, analysing age-at-death distributions from cancers of all sites in Denmark from 1980 to 2014. Data were retrieved from the Danish Cancer Society and the Human Mortality Database.

Results

The main trends captured by PCLM are: (i) a decrease in cancer mortality rates after the 1990s for ages 50–75; (ii) a decrease in cancer mortality in later cohorts for young ages, especially, and very advanced ages. Comparing the raw data by single year of age, with the PCLM-ungrouped distributions, we clearly illustrate that the model fits the data with a high level of accuracy.

Conclusions

The PCLM produces detailed smooth mortality surfaces from death counts observed in coarse age groups with modest assumptions, that is Poisson distributed counts and smoothness of the estimated distribution. Hence, the method has great potential for use within epidemiological research when information is to be gained from aggregated data, because it avoids strict assumptions about the actual distributional shape.

A discrete mixture regression for modeling the duration of non-hospitalization medical leave of motor accident victims

Studies analyzing the temporary repercussions of motor vehicle accidents are scarcer than those analyzing permanent injuries or mortality. A regression model to evaluate the risk factors affecting the duration of temporary disability after injury in such an accident is constructed using a motor insurance dataset. The length of non-hospitalization medical leave, measured in days, following a motor accident is used here as a measure of the severity of temporary disability. The probability function of the number of days of sick leave presents spikes in multiples of five (working week), seven (calendar week) and thirty (month), etc. To account for this, a regression model based on finite mixtures of multiple discrete distributions is proposed to fit the data properly. The model provides a very good fit when the multiples for the working week, week, fortnight and month are taken into account. Victim characteristics of gender and age and accident characteristics of the road user type, vehicle class and the severity of permanent injuries were found to be significant when accounting for the duration of temporary disability.

Tuberculosis Infection in the United States: Prevalence Estimates from the National Health and Nutrition Examination Survey, 2011-2012

BACKGROUND

Reexamining the prevalence of persons infected with tuberculosis (TB) is important to determine trends over time. In 2011-2012 a TB component was included in the National Health and Nutrition Examination Survey (NHANES) to estimate the reservoir of persons infected with TB.

METHODS

Civilian, noninstitutionalized U.S. population survey participants aged 6 years and older were interviewed regarding their TB history and eligibility for the tuberculin skin test (TST) and interferon gamma release assay (IGRA) blood test. Once eligibility was confirmed, both tests were conducted. Prevalence and numbers of TST positive (10 mm or greater), IGRA positive, and both TST and IGRA positive were calculated by adjusting for the complex survey design after applying corrections for item nonresponse and digit preference in TST induration measurements. To examine TST positivity over time, data from NHANES 1999-2000 were reanalyzed using the same statistical methods. The TST was performed using Tubersol, a commercially available purified protein derivative (PPD), rather than PPD-S, which was the antigen used in NHANES 1999-2000. Prior patient history of TB vaccination was not collected in this study nor were patients examined for the presence of a Bacillus of Calmette and Guerin (BCG) vaccine scar.

RESULTS

For NHANES 2011-2012, TST and IGRA results were available for 6,128 (78.4%) and 7,107 (90.9%) eligible participants, respectively. There was no significant difference between the percentage of the U.S. population that was TST positive in 2011-2012 (4.7% [95% CI 3.4-6.3]; 13,276,000 persons) compared with 1999-2000 (4.3%; 3.5-5.3). In 2011-2012 the percentage that was IGRA positive was 5.0% (4.2-5.8) and double TST and IGRA positivity was 2.1% (1.5-2.8). The point estimate of IGRA positivity prevalence in foreign-born persons (15.9%; 13.5-18.7) was lower than for TST (20.5%; 16.1-25.8) in 2011-2012. The point estimate of IGRA positivity prevalence in U.S.-born persons (2.8%; 2.0-3.8) was higher than for TST (1.5%; 0.9-2.6).

CONCLUSIONS

No statistically significant decline in the overall estimated prevalence of TST positivity was detected from 1999-2000 to 2011-2012. The prevalence of TB infection, whether measured by TST or IGRA, remains lower among persons born in the United States compared with foreign-born persons.

Determinants of number-specific recall error of last menstrual period: a retrospective cohort study

OBJECTIVE

To assess the digit preference for last menstrual period (LMP) dates, associated determinants and impact on obstetric outcome.

DESIGN

Retrospective cohort study.

SETTING

University medical centre (the Netherlands).

POPULATION

Cohort of 24 665 LMP records and a subgroup of 4630 cases with known crown-rump length (CRL) measurement, and obstetric outcome.

METHODS

Digit preference was determined by comparing the observed to expected counts of each day. Associated determinants were identified by multivariate regression analysis. Differences in obstetric outcome between LMP and CRL dating were analysed.

MAIN OUTCOME MEASURES

(Non)deprived neighbourhood, cycle irregularity, certainty of LMP date, maternal age, smoking, body mass index, parity and ultrasound investigator. Preterm and post-term delivery.

RESULTS

LMP digit preference for the first [odds ratio (OR), 1.28; 95% confidence interval (95% CI), 1.20-1.36], fifth (OR, 1.10; 95% CI, 1.03-1.17), 10th (OR, 1.17; 95% CI, 1.09-1.25), 15th (OR, 1.31; 95% CI, 1.23-1.40), 20th (OR, 1.22; 95% CI, 1.15-1.30) and 25th (OR, 1.08; 95% CI, 1.01-1.15) days of the month occurred more often than expected. Digit preference occurred more frequently in women living in a deprived neighbourhood (OR, 1.21; 95% CI, 1.06-1.39), with uncertain LMP (OR, 2.03; 95% CI, 1.63-2.52) or irregular cycle (OR, 1.24; 95% CI, 1.06-1.44). More post-term (≥42 weeks) deliveries (OR, 1.27; 95% CI, 1.05-1.54) were observed in LMP dating. This effect was larger in women with a digit preference (OR, 1.56; 95% CI, 1.03-2.37).

CONCLUSIONS

LMP digit preference occurs more often in women living in deprived neighbourhoods, with uncertain LMP or an irregular cycle. LMP-dated pregnancies are associated with more post-term pregnancies.

Correlates of digit bias in self-reporting of cigarette per day (CPD) frequency: results from Global Adult Tobacco Survey (GATS), India and its implications

BACKGROUND

Cigarette per day (CPD) use is a key smoking behaviour indicator. It reflects smoking intensity which is directly proportional to the occurrence of tobacco induced cancers. Self reported CPD assessment in surveys may suffer from digit bias and under reporting. Estimates from such surveys could influence the policy decision for tobacco control efforts. In this context, this study aimed at identifying underlying factors of digit bias and its implications for Global Adult Tobacco Surveillance. MATERIALS OR METHODS: Daily manufactured cigarette users CPD frequencies from Global Adult Tobacco Survey (GATS)--India data were analyzed. Adapted Whipple Index was estimated to assess digit bias and data quality of reported CPD frequency. Digit bias was quantified by considering reporting of '0' or '5' as the terminal digits in the CPD frequency. The factors influencing it were identified by bivariate and logistic regression analysis.

RESULTS

The mean and mode of CPD frequency was 6.7 and 10 respectively. Around 14.5%, 15.1% and 15.2% of daily smokers had reported their CPD frequency as 2, 5 and 10 respectively. Modified Whipple index was estimated to be 226.3 indicating poor data quality. Digit bias was observed in 38% of the daily smokers. Heavy smoking, urban residence, North, South, North- East region of India, less than primary, secondary or higher educated and fourth asset index quintile group were significantly associated with digit bias.

DISCUSSION

The present study highlighted poor quality of CPD frequency data in the GATS-India survey and need for its improvement. Modeling of digit preference and smoothing of the CPD frequency data is required to improve quality of data. Marketing of 10 cigarette sticks per pack may influence CPD frequency reporting, but this needs further examination. Exploring alternative methods to reduce digit bias in cross sectional surveys should be given priority.

A modelling approach for blood units transfused after stem cell transplantation

Patients undergoing stem cell transplantation may require transfusion of units (bags) of packed red blood cells (PRBCs). Modelling of PRBC usage is important not only for prediction of transfusion requirements in future patients but also for its use as an inverse surrogate for engraftment, that is transplantation success. Inspection of PRBC unit usage reveals a strong preference for even numbers, which is caused by behavioural preference on the part of prescribing physicians. Digit preference is a phenomenon observed more commonly with self-reported data: typically survey respondents round recalled quantities such as the age at which a life event occurred to multiples of 5 or 10. In all cases, we can conceive of a latent variable, which has a smooth distribution, which is transformed via stochastic rules to a discrete variable with probability spikes at preferred digits. We propose a modelling framework based on a latent variable specification and stochastic transformation to the spiked distribution. Loglinear models for the mean of the process are implemented. Specification of the stochastic rules is important to success in accurate modelling of the process.

Effects of different manual periodontal probes on periodontal measurements

AIM

To quantify the digit preference effect for three manual periodontal probes and to calculate correction values to enable comparison of studies with equal recording protocols, but different periodontal probes.

MATERIAL AND METHODS

A prospective in vivo crossover study was conducted with a six-sequence three-period design. Six examiners assessed attachment loss (AL), probing pocket depth (PD) and gingiva height (GH) at four surfaces, full-mouth, in six generally healthy subjects using three manual probes: PCP11 (3-3-3-2 mm increments), PCP2 (2 mm increments), and PCPUNC15 (1 mm increments).

RESULTS

Distributions of AL, PD and GH differed between probes (p < 0.001). Compared with PCPUNC15, periodontal measurements coinciding with probe markings of PCP11 and PCP2, respectively, were preferentially named by examiners. Digit preference was most pronounced for PD, but less for AL and GH. In multilevel models, PD differed significantly between all three probes (p < 0.05); probe- and examiner-related effects were also observed for AL and GH. Correction values for pairwise combinations of probes were determined.

CONCLUSIONS

We provided empirical evidence and quantified the effect of probe type on periodontal measurements. Differences in probe type should be considered when comparing periodontal data within and between epidemiological studies and appropriate corrections, provided here, should be applied.

A semi-Markov model for stroke with piecewise-constant hazards in the presence of left, right and interval censoring

This paper presents a parametric method of fitting semi-Markov models with piecewise-constant hazards in the presence of left, right and interval censoring. We investigate transition intensities in a three-state illness–death model with no recovery. We relax the Markov assumption by adjusting the intensity for the transition from state 2 (illness) to state 3 (death) for the time spent in state 2 through a time-varying covariate. This involves the exact time of the transition from state 1 (healthy) to state 2. When the data are subject to left or interval censoring, this time is unknown. In the estimation of the likelihood, we take into account interval censoring by integrating out all possible times for the transition from state 1 to state 2. For left censoring, we use an Expectation–Maximisation inspired algorithm. A simulation study reflects the performance of the method. The proposed combination of statistical procedures provides great flexibility. We illustrate the method in an application by using data on stroke onset for the older population from the UK Medical Research Council Cognitive Function and Ageing Study. Copyright © 2012 John Wiley & Sons, Ltd.