PD, psychological distress; WHO, world health organization; CHD, chronic heart disease; BC, breast cancer; CC, colon cancer; LC, lung cancer; PC, prostate cancer; K6, kessler 6-qusation screening scale; NHIS, national health interview survey; NCHS, national center for health and statistics; CDC, centers for disease control and prevention; AIC, akai information criterion; CO, cumulative odds model; CR, continuation ratio model; AC, adjacent categories model; SAS, statistical analysis system

Cancer is a significant health dilemma in the US and in many countries worldwide. It is one of the leading causes of death around the world. Every year, nearly 10.9 million individuals are diagnosed with cancer worldwide and around 6.7 million individuals die from it. The World Health Organization (WHO) anticipates that death from this disease will continue to rise, with an estimated 11.5 million deaths in 2030.¹ Cancer specialists and medical scientists believe that psychological matters affect the progression of cancer, but the evidence remains unconvincing. Many studies focusing on mental health of cancer patients investigated the psychological impact and stigma of affliction due to cancer.^2–5 Cancer patients naturally experience many kinds of psychological problems due to the physical and mental afflictions caused by the long-term adverse health due to cancer. A number of studies demonstrate a clear link between the affliction from cancer and psychological distress (PD) measured by the symptoms of depression and anxiety.^5–15 PD is a significant problem for patients with cancer at every stage of their disease. Assessment and management of psychological distress are imperative in order to ease the burden on patients and to help them cope with their diagnosis and treatment.¹⁶ A review of the psychological distress literature concludes that there are no simply identifiable characteristics of patients that can readily predict who has the potential need for psychosocial assessment and intervention.^5,6 To explore the impact of PD, many research investigations have been carried out in many countries examining various factors that manipulate the psychological distress among cancer patients. A systematic review on PD among cancer patients conducted by Yeh M et al.⁸ found that the majority of cancer patients face significant psychological and emotional distress at sometime during the course of the illness.

There are various risk factors that have been linked to mental illness. Numerous investigations were conducted to determine the factors that play a significant role on PD among cancer patients.^4-13,17 For example, studies conducted in order to identify the risk factors of PD in females with breast cancer revealed that age, marital status, and education of the patient were the significant factors associated to PD.^11,18,19 Based on prior studies, the socio-demographic and clinical factors of health (i.e., BMI, insurance status, physical activity, smoking, drinking alcohol, etc.) may enhance distress. A common factor that is found to affect PD and cancer was race/ethnicity.^1,3,20 Other studies showed there is correlation between all covariates under consideration and the PD being the response variable. In particular, older women with breast cancer had a significantly higher level of distress Sunderland M et al.¹⁸ These factors can be used to identify individuals who may be at risk of experiencing PD. In this study, our objectives are to determine the prevalence of psychological distress among cancer patients and to investigate the association of psychological distress with various socio-demographic factors.

Although a number of studies investigate PD among cancer patients, the risk factors of PD among the leading sub-types of cancers are not well studied. In this study, we selected sub-types of cancer according to the estimates of the leading new cancer cases and death since 2013 in the US. The most frequently diagnosed cancers that occurred in both males and females in 2013 were breast cancer (BC), colon cancer (CC), lung cancer (LC), and prostate cancer (PC).⁴ In particular, this study explores and compares the prevalence of overall cancer and that off our leading cancer subtypes namely, BC, CC, LC, and PC and the association with psychological distress among the black and white American adults.

Data are obtained from the 2013 National Health Interview Survey(NHIS), which is a large annual survey conducted on a random sample of individuals living in the United States.²² This survey is across-sectional household survey of the US population conducted annually by the National Center for Health Statistics (NCHS), Center for Disease Control and Prevention (CDC). Face-to-face interviews were carried out at respondents’ houses. But, for those who were not home a follow-up was performed over the telephone. This survey employed a randomly selected, stratified, and multistage area design that produced nationally representative sample of households. Data from the NHIS are organized into a number of different types of files. We consider the sample adult module, which contains health-related information on randomly selected adults in the households. According to National Center for Health Statistics during the data collection period in 2013, there were 42, 294 eligible individuals, of which 34, 557 (81.7%) were interviewed.

We consider Kessler6 (K6)^23-26 to measure psychological distress (PD) among cancer patients. This measure K6 contains a six-item instrument, for each part the respondents were asked how frequently they experienced symptoms of psychological distress (sad, nervous, restless, hopeless, everything was an effort, and worthless) during the past 30 days. Each question has a 5-point scale with ranges from 0= ‘none of the time’ to5=‘all of the time’ in the NHIS. Since K6 scores lie between 0 and 4 on the Likert scale, the total of the responses scores is ranged between 0 and 24. We collapsed these scores into three categories as: 0-7 indicating a low level of PD, 8-12 indicating a moderate level of PD, and 13-24 indicating high PD. The predictor variables are selected as those found to be relevant according to the literature review.

Statistical models for ordinal response PD

We consider the three ordinal regression models: cumulative odds (CO), continuation ratio (CR), and adjacent categories (AC) models for psychological distress (PD).^25-27,30 The best-fit model among these three is used to report the statistically significant determinants of PD among cancer patients. The cumulative odds (CO) or the proportional odds (PO) model is appropriate when the response categories are based on one or more latent continuous response variable. A CO model with  categories is divided to  logit equations. We use the category ordering by forming logit s of cumulative probabilities,

$P\left(\raisebox{1ex}{$Y\le j$}\!\left/ \!\raisebox{-1ex}{$x$}\right.\right)={\pi }_1\left(x\right)+\dots +{\pi }_j\left(x\right),j=1,\dots ,J\text{.}$

The cumulative logit s are defined as:

$logit\left[P\left(\raisebox{1ex}{$Y\le j$}\!\left/ \!\raisebox{-1ex}{$x$}\right.\right)\right]=\log \frac{P\left(\raisebox{1ex}{$Y\le j$}\!\left/ \!\raisebox{-1ex}{$x$}\right.\right)}{1-P\left(\raisebox{1ex}{$Y\le j$}\!\left/ \!\raisebox{-1ex}{$x$}\right.\right)}=\log \frac{{\pi }_1\left(x\right)+\dots +{\pi }_j\left(x\right)}{{\pi }_{j+1}\left(x\right)+\dots +{\pi }_{J-1}\left(x\right)},j=1,\dots ,J-1$ (1)

The CO or PO Form of cumulative logit model is:

$logit\left[P\left(\raisebox{1ex}{$Y\le j$}\!\left/ \!\raisebox{-1ex}{$x$}\right.\right)\right]={\alpha }_j+{\beta }^{T}x.$ (2)

Each cumulative logit has its intercept ${\alpha }_j$ . The $\left\{{\text{α}}_{\text{j}}\right\}$ are increasing in $j$ because $\text{P}\left(\raisebox{1ex}{$\text{Y}\le \text{j}$}\!\left/ \!\raisebox{-1ex}{$\text{x}$}\right.\right)$ increases in $j$ for fixed $x$ .For illustration, in our study a three-category outcome will have two binary log it equations based on the following comparisons: 1 vs. 2&3, 1&2 vs. 3. The CO is used to predict the odds of a person being at or below any particular level of Psychological Distress (PD). The following CO model is fitted to our data using the equation form (2):

1. Low vs. (Moderate & High):

$\begin{array}{l}logit\left[P\left(\raisebox{1ex}{$Y\le 1$}\!\left/ \!\raisebox{-1ex}{$x$}\right.\right)\right]=\log \left(\frac{{\pi }_1}{{\pi }_2+{\pi }_3}\right)={\alpha }_1+{\beta }_1{x}_1+{\beta }_2{x}_2+{\beta }_3{x}_3+{\beta }_4{x}_4+{\beta }_5{x}_5+{\beta }_6{x}_6+{\beta }_7{x}_7+\\ {\beta }_8{x}_8+{\beta }_9{x}_9+{\beta }_{10}{x}_{10}+{\beta }_{11}{x}_{11}+{\beta }_{12}{x}_{12}+{\beta }_{13}{x}_{13}+{\beta }_{14}{x}_{14}+{\beta }_{15}{x}_{15}+{\beta }_{16}{x}_{16}+{\beta }_{17}{x}_{17}\end{array}$ (3)

2. (Low & Moderate) vs. High:
$\begin{array}{l}logit\left[P\left(\raisebox{1ex}{$Y\le 2$}\!\left/ \!\raisebox{-1ex}{$x$}\right.\right)\right]=\log \left(\frac{{\pi }_1+{\pi }_2}{{\pi }_3}\right)={\alpha }_2+{\beta }_1{x}_1+{\beta }_2{x}_2+{\beta }_3{x}_3+{\beta }_4{x}_4+{\beta }_5{x}_5+{\beta }_6{x}_6+\\ {\beta }_7{x}_7+{\beta }_8{x}_8+{\beta }_9{x}_9+{\beta }_{10}{x}_{10}+{\beta }_{11}{x}_{11}+{\beta }_{12}{x}_{12}+{\beta }_{13}{x}_{13}+{\beta }_{14}{x}_{14}+{\beta }_{15}{x}_{15}+{\beta }_{16}{x}_{16}+{\beta }_{17}{x}_{17}\end{array}$ (4)

Where ${\text{π}}_1,{\text{π}}_2,{\text{π}}_3$ are the probability of being in low, moderate, and high levels, respectively. ${\text{β}}_1,\dots ,{\text{β}}_{17}$ are the regression coefficients and ${\text{x}}_1,\dots ,{\text{x}}_{17}$ are the covariate factors.

In a continuation ratio (CR) model, the ratio of probabilities as follows,

$\frac{{\pi }_1}{{\pi }_2},\frac{{\pi }_1+{\pi }_2}{{\pi }_3},\dots ,\frac{{\pi }_1+\dots +{\pi }_{J-1}}{{\pi }_J}$

$\frac{{\pi }_1}{{\pi }_2+\dots +{\pi }_J},\frac{{\pi }_2}{{\pi }_3+\dots +{\pi }_J},\dots ,\frac{{\pi }_{J-1}}{{\pi }_J}$

are modeled through “continuation–ratio logit s”. Thus, the model can be written as:

$\log \left(\frac{{\pi }_j}{{\pi }_{j+1}+\dots +{\pi }_J}\right)=x_j^T{\beta }_j$ (5)

The continuation ratio model provides the log odds of the response being in the category j. For our study, J=3, we can estimate the odds of the respondents PD as “Low” vs. “Moderate” and the odds of these levels are in “Low” and “Moderate” versus “High” using:

$\log \left(\frac{{\pi }_1}{{\pi }_2}\right)$ and $\log \left(\frac{{\pi }_1+{\pi }_2}{{\pi }_3}\right)$

The CR may be easier than CO in terms of interpretation if we are interested in finding the probability for individual categories ${\pi }_j$ .

Finally the adjacent categories (AC) model considers the ratios of probabilities for successive categories:

$\frac{{\pi }_1}{{\pi }_2},\frac{{\pi }_2}{{\pi }_3},\dots ,\frac{{\pi }_{J-1}}{{\pi }_J}$

The AC model can be written as:

$\log \left(\frac{{\pi }_j}{{\pi }_{j+1}}\right)=x_j^T{\beta }_j$ (6)

Which is equivalent to ,

$\log \left(\raisebox{1ex}{$P\left(y_i=j\right)$}\!\left/ \!\raisebox{-1ex}{$P\left(y_i=j+1\right)$}\right.\right)=\log \left(\frac{{\pi }_j}{{\pi }_{j+1}}\right)={\beta }_{0j}+{\beta }_1{x}_1+\dots +{\beta }_{p-1}{x}_{p-1}$

This model assumes that the effect of each independent variable to be the same for all adjacent pairs of categories. All computations are conducted using SAS version 9.3 (SAS Institute, Cary, NC,³¹) and the R computing environment (Version3.11, TheRProject). SAS isused to manage the data and create analysis variables. In R, we used the VIGAM package to fit all of the three models.

This study is conducted to determine the prevalence of psychological distress among cancer patients and to investigate the association of psychological distress with various socio-demo graphic factors. Clearly, in terms of mental health, this study has found that cancer patients suffer from higher level of psychological distress than the general population. That is, individuals with cancer diagnostics are more likely to experience higher psychological distress than those without cancer.

In general, the results suggest that psychological distress (PD) is a burden among participants with the cancer, however particular types of cancer is not associated with the level of PD. The results from the ordinal logistic regression models suggest that individuals with self-reported cancer are more likely to experience PD when compared with those without cancer. We found a number of socio-demographic factors that contribute PD other than cancer. These include age, sex, race, marital status, physical activity, BMI, insurance, education level and income. Thus our findings confirm those by the previous findings in the literature. Sub group analysis of PD among the sub types of cancer (breast, colon, lung, and prostate) does not demonstrate any significant determinants of PD.

Psychological distress among cancer patients adds extra burden on them. An important finding is that differential psychological distress level exists a cross different race of overall cancer patients. However, among the different sub types of cancer (breast, colon, lung, and prostate) PD is not found to be different across race.

The psychological distress of cancer patients is not a straight forward subject to study. Numerous factors must be taken into account. However, the main advantage of this study is the availability of all representative population based sample. The NHIS has been conducted every year since1957. However, there are some limitations of this study. One limitation is that the self-reported nature of the data. Another limitation is that a bulk of missing information. There is not enough data to assess the association between the risk factors and the PD across different subtypes of the cancer.