More than 28 million Americans have risk factors and symptoms that could lead to lung cancer, according to analyses of recent medical claims records data by the analysts at Perception Health.
With these predictions and a process to optimize care for lung cancer patients, Perception Health seeks to help hospitals, health plans, employers, and other companies improve the quality of life for people through better healthcare.
According to the American Cancer Society, "lung cancer (both small cell and non-small cell) is the second most common cancer in both men and women (not counting skin cancer). In men, prostate cancer is more common, while in women breast cancer is more common."
There are two main types of lung cancer:
NSCLC is the most common type of lung cancer, affecting 80% of people with lung cancer. This type has three common subtypes:
There are two types of risk factors for NSCLC:
Risk factors you can change
Risk factors you can't change
The American Cancer Society estimates that in 2020 there will be:
The ACS also states on its website that lung cancer mainly occurs in older people. Most people diagnosed with lung cancer are 65 or older; a very small number of people diagnosed are younger than 45. The average age of people when diagnosed is about 70.
However, by using the CARE disease prediction algorithms created by the data scientists at Perception Health, hospitals can identify patients trending towards a lung cancer diagnosis, invite them to lung cancer screenings, and hopefully diagnose them at a much earlier stage. Early identification will often lead to lower costs and better outcomes.
“Our CARE predictive models are machine-learning disease algorithms," says Katie Kruzan, mathematician at Perception Health. "We take our claims dataset, which has about 26 billion rows, and track de-identified patients across the entire continuum of care. We are then able to see what diagnoses, procedure codes, and frequency lead to particular diagnoses, like lung cancer."
Perception Health works with hospitals, ACOs, and other providers to help identify patients who are trending towards specific diseases. "Ultimately, this is all about improving patients' lives, says Kruzan. "We can apply this algorithm to a hospital's patient database to see which of their patients may be trending towards a particular diagnosis, so to encourage those patients to get a screening for that disease."
The sooner you identify a small cancer that has not spread outside the lungs, the better chance you have for treating the disease. If you suspect you may have lung cancer, get a screening as soon as possible.
Symptoms related to Non-Small Cell Lung Cancer (NSCLC) are not specific but can include:
The five basic treatment options for NSCLC are:
The treatment option best suited for you depends upon:
While the number of cancer deaths has dropped steadily over three decades, 2016 and 2017 saw larger drops due to decreases in the number of people smoking and the advancement of new cancer treatments. Still, with early identification from lung cancer screenings, the numbers can drop further. Patients will also benefit from lower costs and better outcomes.
Patients with the risk factors listed above should register for a lung cancer screening and contact their primary care physician for more information on steps to better manage their health.
For the Perception Health study, a sample set of 10,000 patients were chosen with an ICD-10 diagnosis code related to lung cancer rendered by a healthcare provider after January 1, 2016. Any patient that met this criteria had their entire claims history since January 1, 2016 pulled into the dataset. A comparable dataset was created with patients who did not have an ICD-10 diagnosis code related to lung cancer.
The ICD-10 diagnoses and CPT codes that are commonly found in people who had been diagnosed with lung cancer involve respiratory/pulmonary, lung, cardiology, and circulatory symptoms.
The data set used to train the model was comprised of 50% patients (those who had been diagnosed with lung cancer) and 50% non-patients (people without lung cancer related codes in their claims history). The model was then trained on 70% of that data set and then tested on 30% of that same data set.
From the ROC curve, the area under curve (AUC) is calculated as 0.918, and the Random Forest model was accepted with a model confidence value of 0.863.
From the sample of 10,000 medical claims, 27 had a diagnosis of lung cancer in their claims history. After applying the disease prediction algorithm in Perception Health's CARE module, the scientists predicted another 860 patients from the sample are trending toward a lung cancer diagnosis. By projecting these findings nationally, the data scientists estimated that 28 million Americans may be trending towards lung cancer.