High agreement but low kappa: I. The problems of two paradoxes.
ABSTRACT In a fourfold table showing binary agreement of two observers, the observed proportion of agreement, p0, can be paradoxically altered by the chance-corrected ratio that creates kappa as an index of concordance. In one paradox, a high value of p0 can be drastically lowered by a substantial imbalance in the table's marginal totals either vertically or horizontally. In the second pardox, kappa will be higher with an asymmetrical rather than symmetrical imbalanced in marginal totals, and with imperfect rather than perfect symmetry in the imbalance. An adjustment that substitutes kappa max for kappa does not repair either problem, and seems to make the second one worse.
- SourceAvailable from: Ingmar Schäfer[show abstract] [hide abstract]
ABSTRACT: Multimorbidity is a common phenomenon in primary care. Until now, no clinical guidelines for multimorbidity exist. For the development of these guidelines, it is necessary to know whether or not patients are aware of their diseases and to what extent they agree with their doctor. The objectives of this paper are to analyze the agreement of self-reported and general practitioner-reported chronic conditions among multimorbid patients in primary care, and to discover which patient characteristics are associated with positive agreement. The MultiCare Cohort Study is a multicenter, prospective, observational cohort study of 3,189 multimorbid patients, ages 65 to 85. Data was collected in personal interviews with patients and GPs. The prevalence proportions for 32 diagnosis groups, kappa coefficients and proportions of specific agreement were calculated in order to examine the agreement of patient self-reported and general practitioner-reported chronic conditions. Logistic regression models were calculated to analyze which patient characteristics can be associated with positive agreement. We identified four chronic conditions with good agreement (e.g. diabetes mellitus kappa = 0.80;PA = 0,87), seven with moderate agreement (e.g. cerebral ischemia/chronic stroke kappa = 0.55;PA = 0.60), seventeen with fair agreement (e.g. cardiac insufficiency kappa = 0.24;PA = 0.36) and four with poor agreement (e.g. gynecological problems kappa = 0.05;PA = 0.10).Factors associated with positive agreement concerning different chronic diseases were sex, age, education, income, disease count, depression, EQ VAS score and nursing care dependency. For example: Women had higher odds ratios for positive agreement with their GP regarding osteoporosis (OR = 7.16). The odds ratios for positive agreement increase with increasing multimorbidity in almost all of the observed chronic conditions (OR = 1.22-2.41). For multimorbidity research, the knowledge of diseases with high disagreement levels between the patients' perceived illnesses and their physicians' reports is important. The analysis shows that different patient characteristics have an impact on the agreement. Findings from this study should be included in the development of clinical guidelines for multimorbidity aiming to optimize health care. Further research is needed to identify more reasons for disagreement and their consequences in health care.Trial registration: ISRCTN89818205.BMC Family Practice 03/2014; 15(1):39. · 1.61 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Social networking sites such as Facebook have become immensely popular in recent years and present a unique opportunity for researchers to eavesdrop on the collective conversation of current societal issues. We sought to explore doctor-related humor by examining doctor jokes posted on Facebook. We performed a cross-sectional study of 33,326 monitored Facebook users, 263 (0.79%) of whom posted a joke that referenced doctors on their Facebook wall during a 6-month observation period (December 15, 2010 to June 16, 2011). We compared characteristics of so-called jokers to nonjokers and identified the characteristics of jokes that predicted joke success measured by having elicited at least one electronic laugh (eg, an LOL or "laughing out loud") as well as the total number of Facebook "likes" the joke received. Jokers told 156 unique doctor jokes and were the same age as nonjokers but had larger social networks (median Facebook friends 227 vs 132, P<.001) and were more likely to be divorced, separated, or widowed (P<.01). In 39.7% (62/156) of unique jokes, the joke was at the expense of doctors. Jokes at the expense of doctors compared to jokes not at the expense of doctors tended to be more successful in eliciting an electronic laugh (46.5% vs 37.3%), although the association was statistically insignificant. In our adjusted models, jokes that were based on current events received considerably more Facebook likes (rate ratio [RR] 2.36, 95% CI 0.97-5.74). This study provides insight into the use of social networking sites for research pertaining to health and medicine, including the world of doctor-related humor.Journal of Medical Internet Research 01/2014; 16(2):e41. · 3.77 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: The aims of this study were to analyze agreement on information needs within a group of early-state prostate cancer patients and to compare information preferences of patients with the view of health-care professionals about patients' needs. Sample consists of patients (n = 128) and six subgroups of health-care professionals (urologists, n = 32; nurses, n = 95; radiotherapy technologists (RTTs), n = 36; medical oncologists, n = 19; radiation oncologists, n = 12; general practitioners (GPs), n = 10). Information needs have been assessed with 92 questions concerning prostate cancer and its treatment. Respondents judged the importance of addressing each question. Within- and between-group agreements of patients and health-care professional groups were estimated with raw agreement indices as well as chance-corrected Kappa and Gwet's AC1 measures. Finally, group-specific core items rated with high importance as well as high agreement were defined. Patients rated on average (median) half, i.e., 51 out 92 items as essential (interquartile range (IQR) = 36-66), 26 items as desired (IQR = 14-38), and 10 items as avoidable (IQR = 2-22). Within-group agreement on the presented information topics is modest for any participating group (AC1patients = 0.319; AC1professionals = 0.295-0.398). Agreement between patients and professionals is low too (AC1 = 0.282-0.329). Defining group-specific core sets of information topics results in 51 items being part of at least one core set. Concordance of the item core sets of patients and professionals is moderate with κ = 0.38-0.66, sensitivity of professionals' core sets for patients' preferences varies between 56 and 74 %. Results emphasize the need for dialogue between doctor/professional and patient in identifying the information needed by individual patients and support the importance of shared decision making.Supportive Care in Cancer 11/2013; · 2.09 Impact Factor